diff --git a/.github/workflows/CI.yml b/.github/workflows/CI.yml
index 3f209576..4624e2f5 100644
--- a/.github/workflows/CI.yml
+++ b/.github/workflows/CI.yml
@@ -197,10 +197,16 @@ jobs:
           target: ${{ matrix.target }}
           args: --release --out dist --find-interpreter
           manylinux: manylinux_2_28
-          container: "ghcr.io/rust-cross/manylinux_2_28-cross:${{ matrix.target }}"
           before-script-linux: |
-            sudo apt-get update
-            sudo apt-get install --yes --upgrade build-essential cmake protobuf-compiler libssl-dev glibc-source musl-tools
+            # Build images differ per target. The RPM-based image already has the
+            # toolchain and cmake, but its perl lacks the modules openssl-sys needs
+            # to configure a vendored build.
+            if command -v apt-get > /dev/null; then
+              sudo apt-get update
+              sudo apt-get install --yes --upgrade build-essential cmake protobuf-compiler libssl-dev glibc-source musl-tools
+            else
+              yum install -y perl-IPC-Cmd perl-Pod-Html
+            fi
       - name: Upload wheels
         uses: actions/upload-artifact@v4.4.1
         with:
diff --git a/Cargo.lock b/Cargo.lock
index e25c2a68..603aa8b7 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -646,12 +646,14 @@ dependencies = [
 
 [[package]]
 name = "cc"
-version = "1.0.92"
+version = "1.4.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2678b2e3449475e95b0aa6f9b506a28e61b3dc8996592b983695e8ebb58a8b41"
+checksum = "0ad534f4357a5264cce5019c989cf66a4f0dc4e0d1b1d15f8aacec0ff7360273"
 dependencies = [
+ "find-msvc-tools",
  "jobserver",
  "libc",
+ "shlex",
 ]
 
 [[package]]
@@ -726,9 +728,9 @@ checksum = "98cc8fbded0c607b7ba9dd60cd98df59af97e84d24e49c8557331cfc26d301ce"
 
 [[package]]
 name = "cmake"
-version = "0.1.50"
+version = "0.1.58"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a31c789563b815f77f4250caee12365734369f942439b7defd71e18a48197130"
+checksum = "c0f78a02292a74a88ac736019ab962ece0bc380e3f977bf72e376c5d78ff0678"
 dependencies = [
  "cc",
 ]
@@ -1136,6 +1138,12 @@ dependencies = [
  "subtle",
 ]
 
+[[package]]
+name = "find-msvc-tools"
+version = "0.1.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d45db016d36b838f563236e9193d0ee6ce38f3f68b6c94e914b4929c96bbb890"
+
 [[package]]
 name = "flate2"
 version = "1.0.28"
@@ -1672,9 +1680,9 @@ checksum = "e2abad23fbc42b3700f2f279844dc832adb2b2eb069b2df918f455c4e18cc646"
 
 [[package]]
 name = "libc"
-version = "0.2.153"
+version = "0.2.189"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9c198f91728a82281a64e1f4f9eeb25d82cb32a5de251c6bd1b5154d63a8e7bd"
+checksum = "3eaf3ede3fee6db1a4c2ee091bf8a8b4dccdc6d17f656fb07896ee72867612f2"
 
 [[package]]
 name = "libm"
@@ -2690,6 +2698,12 @@ dependencies = [
  "digest",
 ]
 
+[[package]]
+name = "shlex"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f8fadd59c855ef2080decdef8ff161eb6661b86933c9d82e5ba29dc602a55aba"
+
 [[package]]
 name = "signal-hook-registry"
 version = "1.4.1"
diff --git a/configs/dolma3p5-resharding/14t/README.md b/configs/dolma3p5-resharding/14t/README.md
new file mode 100644
index 00000000..df494ccf
--- /dev/null
+++ b/configs/dolma3p5-resharding/14t/README.md
@@ -0,0 +1,125 @@
+# Dolma 3.5 14T materialization
+
+Run from the repository root. Artifacts go to `runs/dolma3p5-resharding/14t/`,
+which is gitignored.
+
+`plan.py` defaults to `inputs/mix-v2.yaml`. The token target is
+`target_uint32_values` in `settings.yaml`. `inputs/README.md` describes the input
+files.
+
+## Requirements
+
+- Python 3.10–3.12
+- `s5cmd` on `PATH`
+- `uv`, to run `materialize.py`
+- `pip install -e .`, for the boto3 and PyYAML imports in `plan.py`,
+  `validate.py`, and `verify.py`
+- AWS credentials: read on the source and destination buckets for plan, preflight,
+  and verify; EC2 create, tag, resume, and pause for `materialize.py --execute`
+
+`materialize.py` runs under `uv` because it needs `poormanray` for the `pmr`
+binary, which is not a repo dependency. The other scripts run under `python`.
+
+All scripts share one build directory. Keep `plan.py --output` and `--build` on
+the others consistent, or leave both at the default.
+
+## 1. Plan
+
+```bash
+python scripts/dolma3p5_resharding/plan.py \
+  --mix configs/dolma3p5-resharding/14t/inputs/mix-v2.yaml \
+  --profile YOUR_READ_ONLY_PROFILE \
+  --destination-root s3://ai2-llm/preprocessed/dolma3p5-14t/materialized \
+  --local-temp-root /mnt/dolma/dolma3p5-resharding \
+  --max-unit-working-bytes 1500000000000
+
+python scripts/dolma3p5_resharding/validate.py
+open runs/dolma3p5-resharding/14t/01-plan/report.html
+```
+
+`--local-temp-root` is worker-local scratch and must be on NVMe.
+`--max-unit-working-bytes` caps the estimated local working set of one execution
+unit; the planner splits units that would exceed it.
+
+Check before continuing:
+
+- source volume, target volume, and sampling rates at every level
+- execution unit working sets fit their instance types
+- destination paths, output shard counts, worker counts
+- `01-plan/execution/validation-summary.json` reports `passed: true`
+- resolution and execution failure CSVs are empty
+
+## 2. Smoke test one category
+
+```bash
+uv run scripts/dolma3p5_resharding/materialize.py --list-categories finemath
+
+uv run scripts/dolma3p5_resharding/materialize.py \
+  --category 'dolma3_finemath_v3:finemath::default' --parallelism 2 --dry-run
+
+uv run scripts/dolma3p5_resharding/materialize.py \
+  --category 'dolma3_finemath_v3:finemath::default' --parallelism 2 \
+  --preflight --execute --verbose
+```
+
+Preflight must pass, every planned unit must dispatch, workers must stop on
+completion, and final verification must pass. Stop if an input drifted or a
+destination is occupied.
+
+## 3. Materialize the rest
+
+Exclude what step 2 materialized. Selectors match by exact string, and an unknown
+selector is an error. `MIX_NAME` excludes every category of that mix;
+`MIX_NAME::CATEGORY_NAME` excludes one. `dolma3_finemath_v3:finemath` has a single
+category, so the mix name covers step 2 exactly.
+
+```bash
+uv run scripts/dolma3p5_resharding/materialize.py \
+  --all --exclude-category 'dolma3_finemath_v3:finemath' \
+  --parallelism 128 --dry-run
+```
+
+Compare the dry run unit count against the plan, then:
+
+```bash
+uv run scripts/dolma3p5_resharding/materialize.py \
+  --all --exclude-category 'dolma3_finemath_v3:finemath' \
+  --parallelism 128 --preflight --execute --verbose
+```
+
+`--parallelism` caps active workers. VM creation runs in batches of 5 with a 3
+second delay; worker setup runs 32 at a time. Each worker takes another compatible
+unit as it finishes one, and stops when its queue is empty. Defaults are cluster
+`dolma3p5-14t`, project `oe-other`, region `us-east-1`. `PMR_REGION` overrides the
+region. Use `--ssh-key-path` if the cluster key is not poormanray's default.
+
+## 4. Verify
+
+```bash
+python scripts/dolma3p5_resharding/verify.py \
+  --category 'dolma3_finemath_v3:finemath::default'
+
+python scripts/dolma3p5_resharding/verify.py --all
+open runs/dolma3p5-resharding/14t/03-output-validation/report.html
+```
+
+Accept only when `output-summary.json` reports `passed: true`, destinations and
+shard counts match the plan, the aggregate residual is within bounds, and the
+problem and error CSVs are empty.
+
+## Validation gates
+
+- `validate.py`: plan consistency, before any worker launches
+- `--preflight`: source drift and destination occupancy for the selected units,
+  immediately before provisioning; a failure blocks the launch and is recorded in
+  `02-preflight/`
+- `materialize.py`: output shard counts, NPY sizes, metadata partners, and
+  residuals, after workers stop
+- `verify.py`: repeats the output checks and writes `03-output-validation/`
+
+## Safety
+
+Plan, preflight, and verification artifacts are disposable. Source objects and
+materialized destinations are not. Materialization refuses occupied destinations
+and never overwrites token or metadata files. Investigate a partially written
+destination rather than retrying into it.
diff --git a/configs/dolma3p5-resharding/14t/inputs/README.md b/configs/dolma3p5-resharding/14t/inputs/README.md
new file mode 100644
index 00000000..5f83bfb6
--- /dev/null
+++ b/configs/dolma3p5-resharding/14t/inputs/README.md
@@ -0,0 +1,33 @@
+# Dolma 3.5 14T plan inputs
+
+A build's id is a hash of the mix and the catalog. Editing either file changes the
+id, invalidates plans already built from it, and changes every destination path
+derived from it. Record notes here rather than in the YAML.
+
+## `mix-v2.yaml`
+
+`plan.py`'s default. 210 top-level mixes, weights summing to 1.0.
+
+Differences from v1:
+
+- `dolma3_finemath_v3:finemath` and `rpj-proofpile-arxiv:proofpile` read
+  decontaminated, ngram-filtered, minhash-deduplicated `dolma2-tokenizer`
+  sources. v1 points both at `dolma3-tokenizer` fallbacks; see its "Missing from
+  our copy" comments.
+- Adds `the-stack-v2:HTML`, weight 0.0137159618, five quality categories.
+- 49 top-level weights differ.
+
+## `dolma3p5-14t-optimal-mix.yaml`
+
+Superseded by v2, retained for provenance. 209 top-level mixes. Materialized once
+on 2026-08-05 to ~14T tokens; output validation failed on 10 of 681 destinations.
+
+## `dolma3p5-reference-all-npy-s3-paths.csv`
+
+Superset of source NPY paths. Planning resolves mix path patterns against this
+catalog instead of listing the bucket, so the catalog is part of the build id.
+
+## Token target
+
+`target_uint32_values` in `../settings.yaml`, 14,000,000,000,000. The mix files
+carry relative weights only.
diff --git a/configs/dolma3p5-resharding/14t/inputs/dolma3p5-14t-optimal-mix.yaml b/configs/dolma3p5-resharding/14t/inputs/dolma3p5-14t-optimal-mix.yaml
new file mode 100644
index 00000000..1364d964
--- /dev/null
+++ b/configs/dolma3p5-resharding/14t/inputs/dolma3p5-14t-optimal-mix.yaml
@@ -0,0 +1,8386 @@
+mix:
+- name: cc_all_dressed/all_dressed_v5:adult_content
+  weight: 0.0003436341
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:art_and_design
+  weight: 0.0056929676
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:crime_and_law
+  weight: 0.0132817202
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:education_and_jobs
+  weight: 0.0129742366
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:electronics_and_hardare
+  weight: 0.0104740419
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:entertainment
+  weight: 0.0290391915
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:fashion_and_beauty
+  weight: 3.85305e-05
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:finance_and_business
+  weight: 0.0245731104
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:food_and_dining
+  weight: 0.0042511664
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:games
+  weight: 0.0208227337
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:health
+  weight: 0.0511882173
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:history_and_geography
+  weight: 0.012334387
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:home_and_hobbies
+  weight: 0.0027664665
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:industrial
+  weight: 0.0071233867
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:literature
+  weight: 0.0309254818
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:politics
+  weight: 0.0098084081
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:religion
+  weight: 0.008963702
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:science_math_and_technology
+  weight: 0.0985427407
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:social_life
+  weight: 0.001391336
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:software
+  weight: 0.014312283
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:software_development
+  weight: 0.0404468733
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:sports_and_fitness
+  weight: 0.005894111
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:transportation
+  weight: 0.0029472001
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:travel_and_tourism
+  weight: 0.0006618033
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:adult_content
+  weight: 1.82661e-05
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:art_and_design
+  weight: 0.0008069679
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:crime_and_law
+  weight: 0.0018826598
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:education_and_jobs
+  weight: 0.0041379178
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:electronics_and_hardare
+  weight: 0.0022270148
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:entertainment
+  weight: 0.0061743793
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:fashion_and_beauty
+  weight: 8.1924e-06
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:finance_and_business
+  weight: 0.0026123955
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:food_and_dining
+  weight: 0.0009038927
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:games
+  weight: 0.0035419018
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:health
+  weight: 0.0057619884
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:history_and_geography
+  weight: 0.001748377
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:home_and_hobbies
+  weight: 0.0007058549
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:industrial
+  weight: 0.0012116726
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:literature
+  weight: 0.0043836311
+  categories:
+  - name: high
+    weight: 0.55
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:politics
+  weight: 0.0013903241
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:religion
+  weight: 0.0012705886
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:science_math_and_technology
+  weight: 0.0145990797
+  categories:
+  - name: high
+    weight: 0.36
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.29
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.04
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:social_life
+  weight: 0.000295829
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:software
+  weight: 0.0030431103
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:software_development
+  weight: 0.0060916006
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:sports_and_fitness
+  weight: 0.0008354796
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:transportation
+  weight: 0.0006266404
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:travel_and_tourism
+  weight: 0.0001407141
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_codetextish:common-pile_github_archive_filtered_decon_ngram_filtered
+  weight: 0.0046144942
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_codetextish/v0/common-pile_github_archive_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_codetextish:common-pile_stackexchange_decon_ngram_filtered
+  weight: 0.0120150982
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_codetextish/v0/common-pile_stackexchange_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_codetextish:common-pile_ubuntu_irc_filtered_decon_ngram_filtered
+  weight: 0.0007835934
+  categories:
+  - name: 100.0000%
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_codetextish/v0/common-pile_ubuntu_irc_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_biodiversity_heritage_library_filtered_decon_ngram_filtered
+  weight: 0.0007289681
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_biodiversity_heritage_library_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_libretexts_filtered_decon_ngram_filtered
+  weight: 6.2483e-05
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_libretexts_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_news_filtered_decon_ngram_filtered
+  weight: 4.16553e-05
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_news_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_oercommons_filtered_decon_ngram_filtered
+  weight: 7.2897e-06
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_oercommons_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_pressbooks_filtered_decon_ngram_filtered
+  weight: 8.33106e-05
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_pressbooks_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_public_domain_review_filtered_decon_ngram_filtered
+  weight: 1.0414e-06
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_public_domain_review_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_youtube_filtered_decon_ngram_filtered
+  weight: 0.0011580179
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_youtube_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_wikish:common-pile_wikimedia_filtered_decon_ngram_filtered
+  weight: 0.0070423439
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_wikish/v0/common-pile_wikimedia_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_wikish:common-pile_wikiteam_filtered_decon_ngram_filtered
+  weight: 0.0014738547
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_wikish/v0/common-pile_wikiteam_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma3_finemath_v3:finemath
+  weight: 0.019975619
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    # Missing from our copy of Dolma 3p5
+    # - dolma3p5_pool/finemath-3plus/finemath/allenai/dolma2-tokenizer/*.npy
+    - preprocessed/dolma3-0625/v0.1-official/allenai/dolma3-tokenizer/finemath-3plus/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:art_and_design
+  weight: 0.0012031016
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.28
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.12
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:crime_and_law
+  weight: 0.0046649874
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.18
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.02
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:education_and_jobs
+  weight: 0.0051531076
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:electronics_and_hardare
+  weight: 0.0030212673
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:entertainment
+  weight: 0.0011801697
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:fashion_and_beauty
+  weight: 2.12938e-05
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:finance_and_business
+  weight: 0.0069147628
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:food_and_dining
+  weight: 0.0005642866
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:games
+  weight: 0.0003562622
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:health
+  weight: 0.0063947018
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:history_and_geography
+  weight: 0.0053824258
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:home_and_hobbies
+  weight: 0.0017698451
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:industrial
+  weight: 0.0028844954
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:literature
+  weight: 0.0049016765
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:politics
+  weight: 0.0060408966
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:religion
+  weight: 0.0041584399
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:science_math_and_technology
+  weight: 0.0408718744
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:social_life
+  weight: 0.0014553516
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:software
+  weight: 0.0009819733
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:software_development
+  weight: 0.0024373249
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:sports_and_fitness
+  weight: 0.0021488344
+  categories:
+  - name: high
+    weight: 0.95
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:transportation
+  weight: 0.002545432
+  categories:
+  - name: high
+    weight: 0.9
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:travel_and_tourism
+  weight: 0.0006846786
+  categories:
+  - name: high
+    weight: 0.55
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:adult_content
+  weight: 0.0002020444
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:art_and_design
+  weight: 0.0004859546
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:crime_and_law
+  weight: 0.0014556969
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:education_and_jobs
+  weight: 0.002722291
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:electronics_and_hardware
+  weight: 0.000295079
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:entertainment
+  weight: 0.0002247564
+  categories:
+  - name: high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:fashion_and_beauty
+  weight: 6.5595e-06
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:finance_and_business
+  weight: 0.0025358673
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:food_and_dining
+  weight: 0.0005921278
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:games
+  weight: 0.0005877942
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:health
+  weight: 0.005828303
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:history_and_geography
+  weight: 0.0031446204
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:home_and_hobbies
+  weight: 0.0003342784
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:industrial
+  weight: 0.0006327061
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:literature
+  weight: 0.0041655781
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:politics
+  weight: 0.0033400265
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:religion
+  weight: 0.0005188506
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:science_math_and_technology
+  weight: 0.0303966842
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:social_life
+  weight: 8.86419e-05
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:software
+  weight: 0.0004571952
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:software_development
+  weight: 0.0024683813
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:sports_and_fitness
+  weight: 0.000152858
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:transportation
+  weight: 0.0002710472
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:travel_and_tourism
+  weight: 0.0003480672
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:agricultural-and-food-sciences
+  weight: 0.0008137318
+  categories:
+  - name: high
+    weight: 0.2727272727272727
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:art
+  weight: 0.0031988383
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:biology
+  weight: 0.0031486849
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:business
+  weight: 0.002134933
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:chemistry
+  weight: 0.0018963333
+  categories:
+  - name: high
+    weight: 0.22
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:computer-science
+  weight: 0.0026756402
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:economics
+  weight: 0.0024845231
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:education
+  weight: 0.0058634982
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:engineering
+  weight: 0.0010131584
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:environmental-science
+  weight: 0.001758634
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:geography
+  weight: 0.0002362256
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:geology
+  weight: 0.0011829084
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:history
+  weight: 0.0038478651
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:law
+  weight: 0.0019550929
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:linguistics
+  weight: 0.0006528847
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:materials-science
+  weight: 0.0007781199
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:mathematics
+  weight: 0.0016844425
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:medicine
+  weight: 0.0008837685
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:philosophy
+  weight: 0.0013942056
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:physics
+  weight: 0.0016464565
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:political-science
+  weight: 0.0052616572
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:psychology
+  weight: 0.0015633621
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:sociology
+  weight: 0.0017592275
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:adult_content
+  weight: 5.47669e-05
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:art_and_design
+  weight: 0.0003734542
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:crime_and_law
+  weight: 0.0033158495
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:education_and_jobs
+  weight: 0.000625632
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:electronics_and_hardare
+  weight: 0.0009109587
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:entertainment
+  weight: 0.001073777
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:fashion_and_beauty
+  weight: 0.0007733966
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:finance_and_business
+  weight: 0.0027681809
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:food_and_dining
+  weight: 0.0009940811
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:games
+  weight: 0.0011503928
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:health
+  weight: 0.0029052424
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:history_and_geography
+  weight: 0.002236105
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:home_and_hobbies
+  weight: 0.0021535216
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:industrial
+  weight: 0.0001684009
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:literature
+  weight: 0.0036108591
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:politics
+  weight: 0.0012073735
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:religion
+  weight: 0.0012412539
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:science_math_and_technology
+  weight: 0.0097818736
+  categories:
+  - name: high
+    weight: 0.19
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.245
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.265
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.09
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:social_life
+  weight: 0.0013101696
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:software
+  weight: 0.0007834838
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:software_development
+  weight: 0.0017355994
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:sports_and_fitness
+  weight: 0.0007919538
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:transportation
+  weight: 0.0005578712
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:travel_and_tourism
+  weight: 0.0011426928
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: rpj-proofpile-arxiv:proofpile
+  weight: 0.0077525178
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    # Missing from our copy of Dolma 3.5 Pool
+    # - dolma3p5_pool/rpj-proofpile-arxiv/proofpile/allenai/dolma2-tokenizer/*.npy
+    - preprocessed/dolma3-0625/v0.1-official/allenai/dolma3-tokenizer/rpj-proofpile-arxiv/*.npy
+    repetition_factor: -1.0
+- name: sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter:sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter
+  weight: 0.0226421906
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Blade
+  weight: 0.0004994127
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Bluespec
+  weight: 4.618e-07
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:C
+  weight: 0.011110817
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:C++
+  weight: 0.0170757764
+  categories:
+  - name: high
+    weight: 0.32
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.27
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:C-Sharp
+  weight: 0.0100775074
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:CSS
+  weight: 0.0015396192
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Clojure
+  weight: 2.7046e-06
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Common_Lisp
+  weight: 8.1139e-06
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Cuda
+  weight: 9.83815e-05
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Dart
+  weight: 0.0004726367
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Erlang
+  weight: 5.4093e-06
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Fortran
+  weight: 0.0002941301
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Fortran_Free_Form
+  weight: 6.76161e-05
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Go
+  weight: 0.0023097669
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Haskell
+  weight: 0.0002434181
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Java
+  weight: 0.030311633
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.24
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:JavaScript
+  weight: 0.01507488
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Java_Server_Pages
+  weight: 0.0015362384
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Julia
+  weight: 7.30254e-05
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Kotlin
+  weight: 0.0011880153
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Lua
+  weight: 0.0012822046
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:MATLAB
+  weight: 0.0011604956
+  categories:
+  - name: high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.24
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Markdown
+  weight: 0.0120486527
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Mathematica
+  weight: 0.0011210754
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:OCaml
+  weight: 0.0001203567
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Objective-C
+  weight: 0.0019973804
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:OpenCL
+  weight: 1.62279e-05
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:PHP
+  weight: 0.0065695828
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Pascal
+  weight: 0.0012738878
+  categories:
+  - name: high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.33
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.33
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Perl
+  weight: 0.0006240968
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Python
+  weight: 0.0167311032
+  categories:
+  - name: high
+    weight: 0.28
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:R
+  weight: 0.0011677305
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:RMarkdown
+  weight: 0.0007478343
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Ruby
+  weight: 0.0016202514
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Rust
+  weight: 0.0010204626
+  categories:
+  - name: high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.23
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:SCSS
+  weight: 0.0001730973
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:SQL
+  weight: 0.0012150618
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Scala
+  weight: 0.0012252042
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Scheme
+  weight: 8.18155e-05
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Shell
+  weight: 0.0012576599
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Swift
+  weight: 0.001224528
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:SystemVerilog
+  weight: 2.7062e-06
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Tcl
+  weight: 0.0001271183
+  categories:
+  - name: high
+    weight: 0.55
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p95/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:TypeScript
+  weight: 0.0028280445
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:VHDL
+  weight: 0.0001687022
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Verilog
+  weight: 0.0006404599
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Vue
+  weight: 0.0009669106
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:jupyter_notebook
+  weight: 0.0026218153
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:reStructuredText
+  weight: 0.0001311753
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: tokyotech-llm_swallow-code-v2:stage5_python_medium_decon_ngram_filtered
+  weight: 0.0237835043
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/tokyotech-llm_swallow-code-v2/stage5_python_medium_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: tokyotech-llm_swallow-math-v2:stage3-qa_decon_ngram_filtered
+  weight: 0.0068098073
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/tokyotech-llm_swallow-math-v2/stage3-qa_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: tokyotech-llm_swallow-math-v2:stage3-texbook_decon_ngram_filtered
+  weight: 0.0091591169
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/tokyotech-llm_swallow-math-v2/stage3-texbook_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
diff --git a/configs/dolma3p5-resharding/14t/inputs/dolma3p5-reference-all-npy-s3-paths.csv b/configs/dolma3p5-resharding/14t/inputs/dolma3p5-reference-all-npy-s3-paths.csv
new file mode 100644
index 00000000..b508ea3b
--- /dev/null
+++ b/configs/dolma3p5-resharding/14t/inputs/dolma3p5-reference-all-npy-s3-paths.csv
@@ -0,0 +1,3918 @@
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_codetextish/v0/common-pile_github_archive_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_codetextish/v0/common-pile_stackexchange_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_codetextish/v0/common-pile_ubuntu_irc_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_biodiversity_heritage_library_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_libretexts_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_news_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_oercommons_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_pressbooks_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_public_domain_review_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_texbookish/v0/common-pile_youtube_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_wikish/v0/common-pile_wikimedia_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/common-pile_wikish/v0/common-pile_wikiteam_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__adult_content/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/finemath/finemath-3plus-decon-sparkle_ngram_filtered_minhash_deduped/allenai/dolma2-tokenizer/0000.csv.gz
+ai2-llm,preprocessed/finemath/finemath-3plus-decon-sparkle_ngram_filtered_minhash_deduped/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/proof-pile-2/v0_decon_v1/documents/rpj-proofpile-arxiv/train_minhash_deduped_ngram_filtered/allenai/dolma2-tokenizer/000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/sponge_211_non-software-development_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C%2B%2B/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p00/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p05/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p10/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p15/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p20/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p25/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p30/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p35/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p40/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p45/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p50/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p55/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p60/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p65/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p70/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p75/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p80/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p85/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p90/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p95/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/tokyotech-llm_swallow-code-v2/stage5_python_medium_decon_ngram_filtered/allenai/dolma2-tokenizer/000.npy
+ai2-llm,preprocessed/tokyotech-llm_swallow-math-v2/stage3-qa_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
+ai2-llm,preprocessed/tokyotech-llm_swallow-math-v2/stage3-texbook_decon_ngram_filtered/allenai/dolma2-tokenizer/0000.npy
\ No newline at end of file
diff --git a/configs/dolma3p5-resharding/14t/inputs/mix-v2.yaml b/configs/dolma3p5-resharding/14t/inputs/mix-v2.yaml
new file mode 100644
index 00000000..708e9ecc
--- /dev/null
+++ b/configs/dolma3p5-resharding/14t/inputs/mix-v2.yaml
@@ -0,0 +1,8425 @@
+mix:
+- name: cc_all_dressed/all_dressed_v5:adult_content
+  weight: 0.0003436341
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/adult_content/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:art_and_design
+  weight: 0.0056929676
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:crime_and_law
+  weight: 0.0132817202
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:education_and_jobs
+  weight: 0.0129742366
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:electronics_and_hardare
+  weight: 0.0104740419
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:entertainment
+  weight: 0.0290391915
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/entertainment/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:fashion_and_beauty
+  weight: 3.85305e-05
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:finance_and_business
+  weight: 0.0245731104
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:food_and_dining
+  weight: 0.0042511664
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:games
+  weight: 0.0208227337
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/games/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:health
+  weight: 0.0511882173
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/health/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:history_and_geography
+  weight: 0.012334387
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:home_and_hobbies
+  weight: 0.0027664665
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:industrial
+  weight: 0.0071233867
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/industrial/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:literature
+  weight: 0.0309254818
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/literature/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:politics
+  weight: 0.0098084081
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/politics/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:religion
+  weight: 0.008963702
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/religion/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:science_math_and_technology
+  weight: 0.0985427407
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:social_life
+  weight: 0.001391336
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/social_life/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:software
+  weight: 0.014312283
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:software_development
+  weight: 0.0404468733
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/software_development/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:sports_and_fitness
+  weight: 0.005894111
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:transportation
+  weight: 0.0029472001
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/transportation/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5:travel_and_tourism
+  weight: 0.0006618033
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5/topic_vigintiles_v2_decon/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:adult_content
+  weight: 1.82661e-05
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:art_and_design
+  weight: 0.0008069679
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:crime_and_law
+  weight: 0.0018826598
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:education_and_jobs
+  weight: 0.0041379178
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:electronics_and_hardare
+  weight: 0.0022270148
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:entertainment
+  weight: 0.0061743793
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:fashion_and_beauty
+  weight: 8.1924e-06
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:finance_and_business
+  weight: 0.0026123955
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:food_and_dining
+  weight: 0.0009038927
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:games
+  weight: 0.0035419018
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:health
+  weight: 0.0057619884
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:history_and_geography
+  weight: 0.001748377
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:home_and_hobbies
+  weight: 0.0007058549
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:industrial
+  weight: 0.0012116726
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:literature
+  weight: 0.0043836311
+  categories:
+  - name: high
+    weight: 0.55
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:politics
+  weight: 0.0013903241
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:religion
+  weight: 0.0012705886
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:science_math_and_technology
+  weight: 0.0145990797
+  categories:
+  - name: high
+    weight: 0.36
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.29
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.04
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:social_life
+  weight: 0.000295829
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:software
+  weight: 0.0030431103
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:software_development
+  weight: 0.0060916006
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:sports_and_fitness
+  weight: 0.0008354796
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:transportation
+  weight: 0.0006266404
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: cc_all_dressed/all_dressed_v5_addons:travel_and_tourism
+  weight: 0.0001407141
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/cc_all_dressed/all_dressed_v5_addons/quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_codetextish:common-pile_github_archive_filtered_decon_ngram_filtered
+  weight: 0.0046144942
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_codetextish/v0/common-pile_github_archive_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_codetextish:common-pile_stackexchange_decon_ngram_filtered
+  weight: 0.0120150982
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_codetextish/v0/common-pile_stackexchange_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_codetextish:common-pile_ubuntu_irc_filtered_decon_ngram_filtered
+  weight: 0.0007835934
+  categories:
+  - name: 100.0000%
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_codetextish/v0/common-pile_ubuntu_irc_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_biodiversity_heritage_library_filtered_decon_ngram_filtered
+  weight: 0.0007289681
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_biodiversity_heritage_library_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_libretexts_filtered_decon_ngram_filtered
+  weight: 6.2483e-05
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_libretexts_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_news_filtered_decon_ngram_filtered
+  weight: 4.16553e-05
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_news_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_oercommons_filtered_decon_ngram_filtered
+  weight: 7.2897e-06
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_oercommons_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_pressbooks_filtered_decon_ngram_filtered
+  weight: 8.33106e-05
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_pressbooks_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_public_domain_review_filtered_decon_ngram_filtered
+  weight: 1.0414e-06
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_public_domain_review_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_texbookish:common-pile_youtube_filtered_decon_ngram_filtered
+  weight: 0.0011580179
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_texbookish/v0/common-pile_youtube_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_wikish:common-pile_wikimedia_filtered_decon_ngram_filtered
+  weight: 0.0070423439
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_wikish/v0/common-pile_wikimedia_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: common-pile_wikish:common-pile_wikiteam_filtered_decon_ngram_filtered
+  weight: 0.0014738547
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/common-pile_wikish/v0/common-pile_wikiteam_filtered_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma3_finemath_v3:finemath
+  weight: 0.019975619
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - preprocessed/finemath/finemath-3plus-decon-sparkle_ngram_filtered_minhash_deduped/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:art_and_design
+  weight: 0.0012031016
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.28
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.12
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__art_and_design/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:crime_and_law
+  weight: 0.0046649874
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.18
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.02
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__crime_and_law/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:education_and_jobs
+  weight: 0.0051531076
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__education_and_jobs/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:electronics_and_hardare
+  weight: 0.0030212673
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__electronics_and_hardare/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:entertainment
+  weight: 0.0011801697
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__entertainment/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:fashion_and_beauty
+  weight: 2.12938e-05
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__fashion_and_beauty/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:finance_and_business
+  weight: 0.0069147628
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__finance_and_business/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:food_and_dining
+  weight: 0.0005642866
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__food_and_dining/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:games
+  weight: 0.0003562622
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__games/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:health
+  weight: 0.0063947018
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__health/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:history_and_geography
+  weight: 0.0053824258
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__history_and_geography/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:home_and_hobbies
+  weight: 0.0017698451
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__home_and_hobbies/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:industrial
+  weight: 0.0028844954
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__industrial/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:literature
+  weight: 0.0049016765
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__literature/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:politics
+  weight: 0.0060408966
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__politics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:religion
+  weight: 0.0041584399
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__religion/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:science_math_and_technology
+  weight: 0.0408718744
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__science_math_and_technology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:social_life
+  weight: 0.0014553516
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__social_life/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:software
+  weight: 0.0009819733
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:software_development
+  weight: 0.0024373249
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__software_development/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:sports_and_fitness
+  weight: 0.0021488344
+  categories:
+  - name: high
+    weight: 0.95
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__sports_and_fitness/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:transportation
+  weight: 0.002545432
+  categories:
+  - name: high
+    weight: 0.9
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__transportation/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/finepdfs_wo_partitioned_qual_ngram_filtered:travel_and_tourism
+  weight: 0.0006846786
+  categories:
+  - name: high
+    weight: 0.55
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/finepdfs_wo_partitioned_qual_ngram_filtered/__label__travel_and_tourism/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:adult_content
+  weight: 0.0002020444
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__adult_content/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:art_and_design
+  weight: 0.0004859546
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__art_and_design/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:crime_and_law
+  weight: 0.0014556969
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__crime_and_law/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:education_and_jobs
+  weight: 0.002722291
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__education_and_jobs/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:electronics_and_hardware
+  weight: 0.000295079
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__electronics_and_hardware/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:entertainment
+  weight: 0.0002247564
+  categories:
+  - name: high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__entertainment/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:fashion_and_beauty
+  weight: 6.5595e-06
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__fashion_and_beauty/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:finance_and_business
+  weight: 0.0025358673
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__finance_and_business/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:food_and_dining
+  weight: 0.0005921278
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__food_and_dining/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:games
+  weight: 0.0005877942
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__games/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:health
+  weight: 0.005828303
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__health/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:history_and_geography
+  weight: 0.0031446204
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__history_and_geography/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:home_and_hobbies
+  weight: 0.0003342784
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__home_and_hobbies/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:industrial
+  weight: 0.0006327061
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__industrial/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:literature
+  weight: 0.0041655781
+  categories:
+  - name: high
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__literature/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:politics
+  weight: 0.0033400265
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__politics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:religion
+  weight: 0.0005188506
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__religion/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:science_math_and_technology
+  weight: 0.0303966842
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__science_math_and_technology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:social_life
+  weight: 8.86419e-05
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__social_life/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:software
+  weight: 0.0004571952
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:software_development
+  weight: 0.0024683813
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__software_development/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:sports_and_fitness
+  weight: 0.000152858
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__sports_and_fitness/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:transportation
+  weight: 0.0002710472
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__transportation/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/olmo-crawled-pdfs_ngram_filtered:travel_and_tourism
+  weight: 0.0003480672
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/olmo-crawled-pdfs_ngram_filtered/__label__travel_and_tourism/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:agricultural-and-food-sciences
+  weight: 0.0008137318
+  categories:
+  - name: high
+    weight: 0.2727272727272727
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.18181818181818182
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__agricultural-and-food-sciences/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:art
+  weight: 0.0031988383
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__art/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:biology
+  weight: 0.0031486849
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__biology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:business
+  weight: 0.002134933
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__business/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:chemistry
+  weight: 0.0018963333
+  categories:
+  - name: high
+    weight: 0.22
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__chemistry/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:computer-science
+  weight: 0.0026756402
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__computer-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:economics
+  weight: 0.0024845231
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__economics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:education
+  weight: 0.0058634982
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__education/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:engineering
+  weight: 0.0010131584
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__engineering/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:environmental-science
+  weight: 0.001758634
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__environmental-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:geography
+  weight: 0.0002362256
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geography/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:geology
+  weight: 0.0011829084
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__geology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:history
+  weight: 0.0038478651
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__history/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:law
+  weight: 0.0019550929
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__law/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:linguistics
+  weight: 0.0006528847
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__linguistics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:materials-science
+  weight: 0.0007781199
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__materials-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:mathematics
+  weight: 0.0016844425
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__mathematics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:medicine
+  weight: 0.0008837685
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__medicine/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:philosophy
+  weight: 0.0013942056
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__philosophy/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:physics
+  weight: 0.0016464565
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__physics/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:political-science
+  weight: 0.0052616572
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__political-science/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:psychology
+  weight: 0.0015633621
+  categories:
+  - name: high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__psychology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: dolma4pdfs/s2orcforolmo_nogpl_ngram_filtered:sociology
+  weight: 0.0017592275
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/dolma4pdfs/dolma4pdfs_full_deduped_partitioned_resharded_qualitytagged_partitioned_decon/s2orcforolmo_nogpl_ngram_filtered/__label__sociology/qual_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:adult_content
+  weight: 5.47669e-05
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/adult_content/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:art_and_design
+  weight: 0.0003734542
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/art_and_design/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:crime_and_law
+  weight: 0.0033158495
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/crime_and_law/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:education_and_jobs
+  weight: 0.000625632
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/education_and_jobs/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:electronics_and_hardare
+  weight: 0.0009109587
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/electronics_and_hardare/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:entertainment
+  weight: 0.001073777
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/entertainment/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:fashion_and_beauty
+  weight: 0.0007733966
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/fashion_and_beauty/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:finance_and_business
+  weight: 0.0027681809
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/finance_and_business/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:food_and_dining
+  weight: 0.0009940811
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/food_and_dining/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:games
+  weight: 0.0011503928
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/games/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:health
+  weight: 0.0029052424
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/health/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:history_and_geography
+  weight: 0.002236105
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/history_and_geography/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:home_and_hobbies
+  weight: 0.0021535216
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/home_and_hobbies/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:industrial
+  weight: 0.0001684009
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/industrial/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:literature
+  weight: 0.0036108591
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/literature/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:politics
+  weight: 0.0012073735
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/politics/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:religion
+  weight: 0.0012412539
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/religion/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:science_math_and_technology
+  weight: 0.0097818736
+  categories:
+  - name: high
+    weight: 0.19
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.245
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.265
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.21
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.09
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/science_math_and_technology/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:social_life
+  weight: 0.0013101696
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.05
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/social_life/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:software
+  weight: 0.0007834838
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:software_development
+  weight: 0.0017355994
+  categories:
+  - name: high
+    weight: 0.45
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/software_development/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:sports_and_fitness
+  weight: 0.0007919538
+  categories:
+  - name: high
+    weight: 0.8
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/sports_and_fitness/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:transportation
+  weight: 0.0005578712
+  categories:
+  - name: high
+    weight: 0.7
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/transportation/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: hplt-project:travel_and_tourism
+  weight: 0.0011426928
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0016/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0017/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0018/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0019/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0012/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0013/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0014/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0015/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0008/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0009/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0010/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0011/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0004/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0005/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0006/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0007/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0000/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0001/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0002/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/hplt-project/three/sorted/eng_Latn/ia_quality_classified_v2_decon_topics_quality_partitioned/travel_and_tourism/vigintile_0003/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: rpj-proofpile-arxiv:proofpile
+  weight: 0.0077525178
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - preprocessed/proof-pile-2/v0_decon_v1/documents/rpj-proofpile-arxiv/train_minhash_deduped_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter:sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter
+  weight: 0.0226421906
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/sponge_211_code_prose_code_prose_tagged_partitioned_decon_ngram_filter/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Blade
+  weight: 0.0004543934
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Blade/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Bluespec
+  weight: 4.202e-07
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Bluespec/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:C
+  weight: 0.0101092375
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:C++
+  weight: 0.0155364883
+  categories:
+  - name: high
+    weight: 0.32
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.27
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C++/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:C-Sharp
+  weight: 0.0091690751
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/C-Sharp/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:CSS
+  weight: 0.0014008309
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/CSS/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Clojure
+  weight: 0.0000024608
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Clojure/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Common_Lisp
+  weight: 0.0000073825
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Common_Lisp/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Cuda
+  weight: 0.0000895129
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Cuda/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Dart
+  weight: 0.0004300311
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Dart/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Erlang
+  weight: 0.0000049217
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Erlang/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Fortran
+  weight: 0.0002676159
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Fortran_Free_Form
+  weight: 0.0000615209
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Fortran_Free_Form/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Go
+  weight: 0.0021015540
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Go/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Haskell
+  weight: 0.0002214753
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Haskell/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:HTML
+  weight: 0.0137159618
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/html/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Java
+  weight: 0.0275792042
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.24
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:JavaScript
+  weight: 0.0137159618
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/JavaScript/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Java_Server_Pages
+  weight: 0.0013977549
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Java_Server_Pages/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Julia
+  weight: 0.0000664426
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Julia/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Kotlin
+  weight: 0.0010809222
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Kotlin/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Lua
+  weight: 0.0011666208
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Lua/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:MATLAB
+  weight: 0.0010558833
+  categories:
+  - name: high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.24
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/MATLAB/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Markdown
+  weight: 0.0109625324
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Markdown/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Mathematica
+  weight: 0.0010200166
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Mathematica/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:OCaml
+  weight: 0.0001095072
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OCaml/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Objective-C
+  weight: 0.0018173275
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Objective-C/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:OpenCL
+  weight: 0.0000147650
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/OpenCL/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:PHP
+  weight: 0.0059773707
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/PHP/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Pascal
+  weight: 0.0011590538
+  categories:
+  - name: high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.33
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.33
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Pascal/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Perl
+  weight: 0.0005678379
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Perl/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Python
+  weight: 0.0152228855
+  categories:
+  - name: high
+    weight: 0.28
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Python/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:R
+  weight: 0.0010624660
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/R/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:RMarkdown
+  weight: 0.0006804211
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/RMarkdown/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Ruby
+  weight: 0.0014741946
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Ruby/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Rust
+  weight: 0.0009284735
+  categories:
+  - name: high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.26
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.23
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Rust/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:SCSS
+  weight: 0.0001574935
+  categories:
+  - name: high
+    weight: 0.66
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.34
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SCSS/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:SQL
+  weight: 0.0011055306
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SQL/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Scala
+  weight: 0.0011147587
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scala/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Scheme
+  weight: 0.0000744403
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Scheme/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Shell
+  weight: 0.0011442887
+  categories:
+  - name: high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Shell/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Swift
+  weight: 0.0011141435
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Swift/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:SystemVerilog
+  weight: 0.0000024623
+  categories:
+  - name: high
+    weight: 0.6
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/SystemVerilog/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Tcl
+  weight: 0.0001156593
+  categories:
+  - name: high
+    weight: 0.55
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p95/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Tcl/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:TypeScript
+  weight: 0.0025731117
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/TypeScript/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:VHDL
+  weight: 0.0001534946
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/VHDL/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Verilog
+  weight: 0.0005827259
+  categories:
+  - name: high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.25
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Verilog/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:Vue
+  weight: 0.0008797489
+  categories:
+  - name: high
+    weight: 0.35
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.15
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/Vue/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:jupyter_notebook
+  weight: 0.0023854730
+  categories:
+  - name: high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.3
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.2
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/jupyter_notebook/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: the-stack-v2:reStructuredText
+  weight: 0.0001193506
+  categories:
+  - name: high
+    weight: 0.5
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p80/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p85/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p90/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p95/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_high
+    weight: 0.4
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p60/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p65/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p70/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p75/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid
+    weight: 0.1
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p40/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p45/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p50/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p55/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: mid_low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p20/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p25/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p30/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p35/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+  - name: low
+    weight: 0.0
+    paths:
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p00/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p05/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p10/allenai/dolma2-tokenizer/*.npy
+    - dolma3p5_pool/the-stack-v2/spring2code_v2/minhash_filter_v2_2026_stack_edu_redux_tagged_decon_reshard_partitioned_gzip_ngram_filtered/reStructuredText/quality_p15/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: tokyotech-llm_swallow-code-v2:stage5_python_medium_decon_ngram_filtered
+  weight: 0.0237835043
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/tokyotech-llm_swallow-code-v2/stage5_python_medium_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: tokyotech-llm_swallow-math-v2:stage3-qa_decon_ngram_filtered
+  weight: 0.0068098073
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/tokyotech-llm_swallow-math-v2/stage3-qa_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
+- name: tokyotech-llm_swallow-math-v2:stage3-texbook_decon_ngram_filtered
+  weight: 0.0091591169
+  categories:
+  - name: default
+    weight: 1.0
+    paths:
+    - dolma3p5_pool/tokyotech-llm_swallow-math-v2/stage3-texbook_decon_ngram_filtered/allenai/dolma2-tokenizer/*.npy
+    repetition_factor: -1.0
diff --git a/configs/dolma3p5-resharding/14t/settings.yaml b/configs/dolma3p5-resharding/14t/settings.yaml
new file mode 100644
index 00000000..e6305fe6
--- /dev/null
+++ b/configs/dolma3p5-resharding/14t/settings.yaml
@@ -0,0 +1,63 @@
+# Preparation-only defaults for the Dolma 3.5 14T mix.
+#
+# This file contains no destination. A new, sufficiently deep S3 destination
+# must be supplied explicitly to the `plan.py` command.
+target_uint32_values: 14000000000000
+direct_s3_bucket: ai2-llm
+
+# Inventory concurrency. Inventory operations are read-only.
+s5cmd_numworkers: 64
+s5cmd_retry_count: 10
+inventory_max_workers: 32
+
+# Upper bound on target / available tokens per leaf. Unset means the plan records
+# each leaf's sampling rate and reports `not_checked`. A value above 1.0 makes the
+# inventory stage flag leaves whose rate exceeds it.
+# maximum_expected_upsample_rate:
+
+# Catalog-backed objects are grouped under narrow common prefixes. Groups that
+# exceed either limit are split into smaller listings.
+max_listing_catalog_objects: 5000
+max_listing_overfetch_ratio: 8.0
+minimum_catalog_prefix_components: 4
+
+# Deterministic per-execution-unit config values. Worker storage is deliberately
+# not defaulted here; plan.py requires an explicit planning budget.
+random_seed: 42
+max_workers_per_reshard: 32
+s5cmd_download_concurrency: 32
+tokenizer_name_or_path: allenai/dolma2-tokenizer
+
+# Choose enough output shards to keep their average size near 64 GiB, but never
+# create more than eight outputs for one execution unit. Units smaller than the
+# target produce one shard.
+target_output_shard_bytes: 68719476736
+max_output_shards_per_unit: 8
+
+# Each unit is assigned the smallest worker satisfying both its estimated CPU
+# work and its peak local working set. CPU work includes planned output plus
+# the two full metadata passes required for every document-selected source.
+# Worker selection does not increase the output shard count.
+worker_disk_headroom_ratio: 1.1
+document_selection_work_passes: 2
+worker_instance_grid:
+  - instance_type: i4i.2xlarge
+    vcpus: 8
+    local_nvme_devices: 1
+    local_nvme_bytes: 1875000000000
+    max_estimated_work_uint32_values: 150000000000
+  - instance_type: i4i.4xlarge
+    vcpus: 16
+    local_nvme_devices: 1
+    local_nvme_bytes: 3750000000000
+    max_estimated_work_uint32_values: 300000000000
+  - instance_type: i4i.8xlarge
+    vcpus: 32
+    local_nvme_devices: 2
+    local_nvme_bytes: 7500000000000
+    max_estimated_work_uint32_values:
+
+# Partial copies are selected on whole-document boundaries while materializing.
+# These bounds are checked from output file sizes after all units finish.
+max_materialized_unit_target_residual_fraction: 0.001
+max_materialized_total_target_residual_fraction: 0.00001
diff --git a/pyproject.toml b/pyproject.toml
index 2dd5ad58..0e43e6a5 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -247,3 +247,10 @@ plugins = ["numpy.typing.mypy_plugin"]
 
 [tool.mypy-tests]
 strict_optional = false
+
+[tool.pytest.ini_options]
+# Tests under tests/python import helper packages from scripts/ (for example
+# scripts.resharding.dispatch). Bare `pytest tests/python/` does not put the
+# repository root on sys.path, so make it explicit instead of depending on the
+# invocation directory.
+pythonpath = ["."]
diff --git a/python/dolma/taggers/quality.py b/python/dolma/taggers/quality.py
index 76f5b6d2..b864fb1b 100644
--- a/python/dolma/taggers/quality.py
+++ b/python/dolma/taggers/quality.py
@@ -30,7 +30,7 @@ def predict_slice(self, text_slice: TextSlice) -> Iterable[Prediction]:
         pred = self.classifier.predict(text)
 
         # Extract the predicted label and its probability
-        (pred_label, pred_prob) = pred
+        pred_label, pred_prob = pred
         pred_label = pred_label[0]
         probability_score = pred_prob[0]
 
diff --git a/python/dolma/tokenizer/document_selection.py b/python/dolma/tokenizer/document_selection.py
new file mode 100644
index 00000000..898945f5
--- /dev/null
+++ b/python/dolma/tokenizer/document_selection.py
@@ -0,0 +1,259 @@
+"""Sample whole documents from a token memmap at document boundaries.
+
+Reads the metadata paired with a memmap and writes a metadata index describing
+the selected documents. Planners still derive total token counts from the memmap
+size; this module is only used when materializing a partial copy.
+
+Each document is keyed on (seed, start, end, document_id, source,
+source_index), joined with NUL separators and hashed with blake2b
+(digest_size=8, person=b"dolma-doc-v1"). The leading bits of the digest assign
+the document to one of DOCUMENT_HASH_BUCKETS buckets, and buckets are taken in
+ascending order until the requested token count is reached. File name, row
+order, and worker count do not affect the result.
+
+Because start and end are token offsets into one specific memmap, a selection
+reproduces only against byte-identical source metadata. Re-tokenizing,
+reshuffling, or re-concatenating a source changes those offsets, so the same
+seed selects a different set of documents and nothing in the inputs reports
+that the source moved. Manifests record source sizes and ETags to pin this.
+
+DOCUMENT_SELECTION_ALGORITHM is the version string planners pin against.
+Changing the keying, the hash, or the bucket walk requires bumping it so a
+manifest requesting the old algorithm is rejected.
+"""
+
+from __future__ import annotations
+
+import csv
+import gzip
+import hashlib
+import time
+from collections.abc import Callable, Iterator
+from dataclasses import dataclass
+from pathlib import Path
+
+import smart_open
+
+DOCUMENT_SELECTION_ALGORITHM = "document_hash_bucket_v1"
+DOCUMENT_HASH_BUCKETS = 1 << 16
+
+# Digest width in bits, derived from the blake2b digest size rather than
+# restated, since _hash_bucket shifts a digest down to its leading bucket bits.
+_HASH_DIGEST_SIZE_BYTES = 8
+_HASH_BITS = _HASH_DIGEST_SIZE_BYTES * 8
+_HASH_BUCKET_BITS = DOCUMENT_HASH_BUCKETS.bit_length() - 1
+
+
+@dataclass(frozen=True)
+class DocumentMetadataRow:
+    row_number: int
+    start: int
+    end: int
+    document_id: str
+    source: str
+    source_index: int
+
+    @property
+    def token_count(self) -> int:
+        return self.end - self.start
+
+    def as_csv_row(self) -> list[str | int]:
+        return [
+            self.start,
+            self.end,
+            self.document_id,
+            self.source,
+            self.source_index,
+        ]
+
+
+@dataclass(frozen=True)
+class DocumentSelectionResult:
+    selection_path: Path
+    requested_uint32_values: int
+    selected_uint32_values: int
+    target_residual_uint32_values: int
+    selected_document_count: int
+    source_document_count: int
+    largest_document_uint32_values: int
+
+
+def _iter_metadata_rows(metadata_path: Path, source_uint32_values: int) -> Iterator[DocumentMetadataRow]:
+    previous_end = 0
+    with smart_open.open(metadata_path, "r", encoding="utf-8") as handle:
+        reader = csv.reader(handle)
+        for row_number, row in enumerate(reader, start=1):
+            if len(row) != 5:
+                raise ValueError(
+                    f"Invalid token metadata row {metadata_path}:{row_number}; " "expected five columns"
+                )
+            try:
+                start = int(row[0])
+                end = int(row[1])
+                source_index = int(row[4])
+            except ValueError as exc:
+                raise ValueError(f"Invalid token offsets on {metadata_path}:{row_number}") from exc
+            if start != previous_end or end <= start or end > source_uint32_values:
+                raise ValueError(
+                    f"Non-contiguous token metadata on {metadata_path}:{row_number}; "
+                    f"expected start {previous_end}, found [{start}, {end})"
+                )
+            previous_end = end
+            yield DocumentMetadataRow(
+                row_number=row_number,
+                start=start,
+                end=end,
+                document_id=row[2],
+                source=row[3],
+                source_index=source_index,
+            )
+    if previous_end != source_uint32_values:
+        raise ValueError(
+            f"Token metadata does not cover its memmap: {metadata_path}; "
+            f"covered {previous_end} of {source_uint32_values} uint32 values"
+        )
+
+
+def _document_hash(row: DocumentMetadataRow, seed: int) -> int:
+    """Hash the document identity tuple with the seed mixed into the payload."""
+
+    payload = (f"{seed}\0{row.start}\0{row.end}\0{row.document_id}\0" f"{row.source}\0{row.source_index}").encode(
+        "utf-8"
+    )
+    return int.from_bytes(
+        hashlib.blake2b(
+            payload,
+            digest_size=_HASH_DIGEST_SIZE_BYTES,
+            person=b"dolma-doc-v1",
+        ).digest(),
+        "big",
+    )
+
+
+def _hash_bucket(hash_value: int) -> int:
+    """Take the leading `_HASH_BUCKET_BITS` of a `_HASH_BITS`-wide digest."""
+
+    return hash_value >> (_HASH_BITS - _HASH_BUCKET_BITS)
+
+
+def create_document_selection(
+    *,
+    metadata_path: str | Path,
+    selection_path: str | Path,
+    source_uint32_values: int,
+    target_uint32_values: int,
+    seed: int,
+    progress: Callable[[str, int, int], None] | None = None,
+    progress_interval_seconds: float = 10.0,
+) -> DocumentSelectionResult:
+    """Select whole documents up to target_uint32_values, ranked by hash bucket.
+
+    The first pass totals document lengths per bucket. The second pass writes
+    rows below the threshold bucket to the selection index, holding only the
+    threshold bucket in memory. The selected count differs from the requested
+    count by at most the largest document in the source.
+
+    Writes a gzipped CSV to selection_path and fails if it already exists. See
+    the module docstring for what the result depends on.
+    """
+
+    metadata_path = Path(metadata_path)
+    selection_path = Path(selection_path)
+    if source_uint32_values <= 0:
+        raise ValueError("source_uint32_values must be positive")
+    if not 0 < target_uint32_values < source_uint32_values:
+        raise ValueError("target_uint32_values must be between zero and the source size")
+    if selection_path.exists() or selection_path.is_symlink():
+        raise FileExistsError(f"Refusing to replace selection index: {selection_path}")
+
+    bucket_values = [0] * DOCUMENT_HASH_BUCKETS
+    source_document_count = 0
+    largest_document = 0
+    last_progress_at = time.monotonic()
+    if progress is not None:
+        progress("pass 1/2", 0, 0)
+    for row in _iter_metadata_rows(metadata_path, source_uint32_values):
+        bucket_values[_hash_bucket(_document_hash(row, seed))] += row.token_count
+        source_document_count += 1
+        largest_document = max(largest_document, row.token_count)
+        if progress is not None and time.monotonic() - last_progress_at >= progress_interval_seconds:
+            progress("pass 1/2", source_document_count, row.end)
+            last_progress_at = time.monotonic()
+    if progress is not None:
+        progress("pass 1/2", source_document_count, source_uint32_values)
+
+    below_threshold_values = 0
+    threshold_bucket = -1
+    threshold_target = 0
+    for bucket, bucket_total in enumerate(bucket_values):
+        if below_threshold_values + bucket_total < target_uint32_values:
+            below_threshold_values += bucket_total
+            continue
+        threshold_bucket = bucket
+        threshold_target = target_uint32_values - below_threshold_values
+        break
+    if threshold_bucket < 0:
+        raise RuntimeError("Could not locate the document-selection hash threshold")
+
+    selected_documents = 0
+    threshold_rows: list[tuple[int, DocumentMetadataRow]] = []
+    selection_path.parent.mkdir(parents=True, exist_ok=True)
+    second_pass_documents = 0
+    last_progress_at = time.monotonic()
+    if progress is not None:
+        progress("pass 2/2", 0, 0)
+    with gzip.open(selection_path, "xt", encoding="utf-8", newline="") as handle:
+        writer = csv.writer(handle)
+        for row in _iter_metadata_rows(metadata_path, source_uint32_values):
+            second_pass_documents += 1
+            hash_value = _document_hash(row, seed)
+            bucket = _hash_bucket(hash_value)
+            if bucket < threshold_bucket:
+                writer.writerow(row.as_csv_row())
+                selected_documents += 1
+            elif bucket == threshold_bucket:
+                threshold_rows.append((hash_value, row))
+            if progress is not None and time.monotonic() - last_progress_at >= progress_interval_seconds:
+                progress("pass 2/2", second_pass_documents, row.end)
+                last_progress_at = time.monotonic()
+
+        if progress is not None:
+            progress("pass 2/2", second_pass_documents, source_uint32_values)
+
+        threshold_rows.sort(
+            key=lambda item: (
+                item[0],
+                item[1].start,
+                item[1].end,
+                item[1].row_number,
+            )
+        )
+        selected_threshold_values = 0
+        best_error = threshold_target
+        for _, row in threshold_rows:
+            candidate_values = selected_threshold_values + row.token_count
+            candidate_error = abs(threshold_target - candidate_values)
+            if candidate_error > best_error:
+                break
+            writer.writerow(row.as_csv_row())
+            selected_documents += 1
+            selected_threshold_values = candidate_values
+            best_error = candidate_error
+
+    selected_values = below_threshold_values + selected_threshold_values
+
+    residual = selected_values - target_uint32_values
+    if abs(residual) > largest_document:
+        raise RuntimeError(
+            "Document selection exceeded its one-document residual bound: "
+            f"{residual} versus largest document {largest_document}"
+        )
+    return DocumentSelectionResult(
+        selection_path=selection_path,
+        requested_uint32_values=target_uint32_values,
+        selected_uint32_values=selected_values,
+        target_residual_uint32_values=residual,
+        selected_document_count=selected_documents,
+        source_document_count=source_document_count,
+        largest_document_uint32_values=largest_document,
+    )
diff --git a/python/dolma/tokenizer/reshard.py b/python/dolma/tokenizer/reshard.py
index 69a50b8a..1d5b2a57 100644
--- a/python/dolma/tokenizer/reshard.py
+++ b/python/dolma/tokenizer/reshard.py
@@ -1,37 +1,19 @@
-"""
-# Adding tool to reshard npy files based on minimum desired size.
-
-Given a prefix with npy and csv.gz files, this script will merge the npy files so that the output
-satisfies a minimum size constraint.
-
-
-## Usage
-
-In case we wanna reshard from S3, we can do:
-
-```bash
-python -m dolma.tokenizer.reshard -s s3://bucket/prefix -d s3://bucket/prefix-resharded
-```
-
-If you wanna customize which local tempdir to use, you can do:
-
-```bash
-python -m dolma.tokenizer.reshard -s s3://bucket/prefix -d s3://bucket/prefix-resharded -l /mnt/raid0/tempdir
-```
-
-If you want to reshard locally, you can do:
+"""Reshard npy token files into larger output shards.
 
-```
-python -m dolma.tokenizer.reshard -s /path/to/local/prefix -d /path/to/local/prefix-resharded
-```
+Merges npy files and their csv.gz metadata partners so that each output shard
+satisfies a size or file-count constraint. The entry point takes a single YAML
+config path, or "-" to read the config from stdin.
 
-To change number of workers, you can do:
+A config names one destination and at least one source, and may mix the two source
+kinds. source_prefixes discovers every npy under a prefix by listing it.
+source_manifests reads an exact CSV manifest of npy/csv.gz pairs, with per-row
+repeat_count and deterministic partial selection, so a run can reproduce an exact
+subset of tokens rather than whole files. Exactly one of max_size_bytes or
+max_num_files must be set; ReshardingConfig holds the remaining fields.
 
-```bash
-python -m dolma.tokenizer.reshard -s s3://bucket/prefix -d s3://bucket/prefix-resharded -w 10
-```
-
-## Contact info
+Destinations are never overwritten: reshard() refuses to run when the destination
+already contains objects, and token and metadata files are written with no-clobber
+semantics.
 
 Author: Luca Soldaini
 Email:  luca@soldaini.net
@@ -44,114 +26,366 @@
 import os
 import random
 import re
+import shlex
 import shutil
 import subprocess
 import sys
 import threading
+import time
 from collections import Counter
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from dataclasses import asdict, dataclass
+from collections.abc import Callable, Mapping
+from concurrent.futures import (
+    FIRST_COMPLETED,
+    ProcessPoolExecutor,
+    ThreadPoolExecutor,
+    as_completed,
+    wait,
+)
+from dataclasses import asdict, dataclass, field
 from functools import partial
 from pathlib import Path
 from tempfile import mkdtemp
 from urllib.parse import urlparse
 
+import boto3
 import numpy as np
 import smart_open
 import yaml
-from tqdm import tqdm
 
 from dolma.core.loggers import get_logger
+from dolma.tokenizer.document_selection import (
+    DOCUMENT_SELECTION_ALGORITHM,
+    DocumentSelectionResult,
+    create_document_selection,
+)
 from dolma.tokenizer.tokenizer import Tokenizer
 
 logger = get_logger(__name__)
 logger.setLevel(logging.INFO)
 
+# Manifests carry no version column. External planners import this constant and
+# refuse to dispatch when it does not match the version they generated for. Bump it
+# when the columns the parser requires, or their meaning, change.
+RESHARDING_MANIFEST_SCHEMA_VERSION = 2
+PROGRESS_INTERVAL_SECONDS = 10.0
+# Default parallelism for a single s5cmd `cp`, shared by the manifest downloader
+# and the config field that feeds it.
+DEFAULT_S5CMD_DOWNLOAD_CONCURRENCY = 32
+# Rows between progress checks while writing merged metadata.
+METADATA_ROW_LOG_INTERVAL = 100_000
+
+
+def _human_count(value: int, unit: str = "") -> str:
+    for scale, suffix in ((10**12, "T"), (10**9, "B"), (10**6, "M"), (10**3, "K")):
+        if abs(value) >= scale:
+            rendered = f"{value / scale:.3g}{suffix}"
+            return f"{rendered} {unit}".rstrip()
+    return f"{value:,} {unit}".rstrip()
+
+
+def _human_bytes(value: int) -> str:
+    for scale, suffix in (
+        (1024**4, "TiB"),
+        (1024**3, "GiB"),
+        (1024**2, "MiB"),
+        (1024, "KiB"),
+    ):
+        if value >= scale:
+            return f"{value / scale:.3g} {suffix}"
+    return f"{value:,} B"
+
+
+def _elapsed(started_at: float) -> str:
+    seconds = max(0, round(time.monotonic() - started_at))
+    minutes, seconds = divmod(seconds, 60)
+    hours, minutes = divmod(minutes, 60)
+    if hours:
+        return f"{hours}h {minutes:02d}m {seconds:02d}s"
+    if minutes:
+        return f"{minutes}m {seconds:02d}s"
+    return f"{seconds}s"
+
+
+def _run_s5cmd(command: list[str], phase: str) -> None:
+    """Run s5cmd, forwarding its in-place progress updates to the worker log."""
+
+    logger.info("s5cmd %s: %s", phase, shlex.join(command))
+    process = subprocess.Popen(
+        command,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+        encoding="utf-8",
+        errors="replace",
+        bufsize=1,
+    )
+    assert process.stdout is not None
+    for line in process.stdout:
+        # TextIOWrapper's universal-newline handling converts the carriage returns
+        # used by --show-progress into newlines. Flush each update so the controller
+        # can retrieve it while s5cmd is still running.
+        print(line, end="", flush=True)
+    returncode = process.wait()
+    if returncode:
+        raise RuntimeError(f"s5cmd {phase} failed with exit code {returncode}; inspect the worker log")
+
 
 @dataclass(frozen=True)
 class TokensMetadataPaths:
     npy_path: str
     csv_path: str
+    selection_path: str | None = None
+    selected_uint32_values: int | None = None
 
     def __post_init__(self):
-        assert self.npy_path.endswith(".npy")
-        assert self.csv_path.endswith(".csv.gz")
-        assert Path(self.npy_path).stem == Path(Path(self.csv_path).stem).stem
+        # These checks raise rather than assert because asserts are stripped
+        # under `python -O`.
+        if not self.npy_path.endswith(".npy"):
+            raise ValueError(f"Token path must end in .npy: {self.npy_path}")
+        if not self.csv_path.endswith(".csv.gz"):
+            raise ValueError(f"Metadata path must end in .csv.gz: {self.csv_path}")
+        if Path(self.npy_path).stem != Path(Path(self.csv_path).stem).stem:
+            raise ValueError(f"Mismatched token/metadata pair: {self.npy_path}, {self.csv_path}")
+        if self.selection_path is None:
+            if self.selected_uint32_values is not None:
+                raise ValueError("selected_uint32_values requires a selection_path")
+        else:
+            if not self.selection_path.endswith(".csv.gz"):
+                raise ValueError(f"Selection path must end in .csv.gz: {self.selection_path}")
+            if self.selected_uint32_values is None:
+                raise ValueError("A selection_path requires selected_uint32_values")
+            if self.selected_uint32_values <= 0:
+                raise ValueError(f"selected_uint32_values must be positive: {self.selected_uint32_values}")
 
     @property
     def size(self) -> int:
+        if self.selected_uint32_values is not None:
+            return self.selected_uint32_values * np.dtype(np.uint32).itemsize
         return os.path.getsize(self.npy_path)
 
 
+@dataclass(frozen=True)
+class MergeGroupResult:
+    destination: Path
+    output_bytes: int
+    metadata_bytes: int
+    document_count: int
+    token_copy_operations: int
+    elapsed_seconds: float
+
+
+@dataclass(frozen=True)
+class MergeGroupProgress:
+    destination: Path
+    processed_values: int
+    total_values: int
+    document_count: int
+    complete: bool = False
+
+
 def merge_group(
     paths: list[TokensMetadataPaths],
     destination: str | Path,
     dtype: np.dtype,
-):
-    """
-    Given a list of paths, merge them into a single memmap.
+    progress: Callable[[MergeGroupProgress], None] | None = None,
+) -> MergeGroupResult:
+    """Merge the given inputs into one npy memmap plus a sibling csv.gz of metadata.
+
+    Document offsets in the merged metadata are rebased onto the output memmap. Inputs
+    carrying a selection index contribute only their selected documents and require a
+    uint32 dtype. Raises FileExistsError when either output path exists, and
+    RuntimeError when the number of values written does not match the total computed
+    from the inputs.
     """
     npy_destination = Path(destination)
     csv_destination = npy_destination.with_suffix(".csv.gz")
     total_size = sum(p.size for p in paths)
+    total_values = total_size // dtype.itemsize
+    if any(path.selection_path is not None for path in paths) and dtype.itemsize != 4:
+        raise ValueError("Document selections require uint32 token memmaps")
 
     npy_destination.parent.mkdir(parents=True, exist_ok=True)
+    if os.path.lexists(npy_destination) or os.path.lexists(csv_destination):
+        raise FileExistsError(f"Refusing to replace an existing reshard output: {npy_destination}")
 
-    target_memmap = np.memmap(npy_destination, mode="w+", shape=(total_size // dtype.itemsize,), dtype=dtype)
+    started_at = time.monotonic()
+    logger.info(
+        "merge %s start · %s inputs · %s tokens (%s)",
+        npy_destination.stem,
+        len(paths),
+        _human_count(total_values),
+        _human_bytes(total_size),
+    )
+    target_memmap = np.memmap(npy_destination, mode="w+", shape=(total_values,), dtype=dtype)
+
+    token_offset = 0
+    completed_values = 0
+    document_count = 0
+    token_copy_operations = 0
+    last_progress_at = started_at
+
+    def publish_progress(*, complete: bool = False) -> None:
+        if progress is not None:
+            progress(
+                MergeGroupProgress(
+                    destination=npy_destination,
+                    processed_values=completed_values + input_processed_values,
+                    total_values=total_values,
+                    document_count=document_count,
+                    complete=complete,
+                )
+            )
 
-    bytes_offset = row_offset = 0
+    input_processed_values = 0
+    publish_progress()
     with smart_open.open(csv_destination, "w", encoding="utf-8") as f:
-        for path in paths:
-            rw = csv.writer(f)
-            source_memmap = np.memmap(path.npy_path, mode="r", dtype=dtype, shape=(path.size // dtype.itemsize,))
-            target_memmap[bytes_offset : bytes_offset + source_memmap.shape[0]] = source_memmap
-            target_memmap.flush()
-
-            row_count = 0
-            with smart_open.open(path.csv_path, "r", encoding="utf-8") as g:
+        rw = csv.writer(f)
+        for input_index, path in enumerate(paths, start=1):
+            input_started_at = time.monotonic()
+            input_documents = 0
+            input_processed_values = 0
+            view_type = "selected" if path.selection_path is not None else "full"
+            source_memmap = np.memmap(
+                path.npy_path,
+                mode="r",
+                dtype=dtype,
+                shape=(os.path.getsize(path.npy_path) // dtype.itemsize,),
+            )
+            metadata_path = path.selection_path or path.csv_path
+            if path.selection_path is None:
+                copy_started_at = time.monotonic()
+                target_memmap[token_offset : token_offset + source_memmap.shape[0]] = source_memmap
+                copy_seconds = max(time.monotonic() - copy_started_at, 1e-9)
+                token_copy_operations += 1
+                logger.info(
+                    "merge %s copy · input %s/%s full · %s at %s/s",
+                    npy_destination.stem,
+                    input_index,
+                    len(paths),
+                    _human_bytes(path.size),
+                    _human_bytes(round(path.size / copy_seconds)),
+                )
+                copy_start = token_offset
+                token_offset += source_memmap.shape[0]
+            else:
+                copy_start = None
+            with smart_open.open(metadata_path, "r", encoding="utf-8") as g:
                 rd = csv.reader(g)
                 for row in rd:
                     start, end, id_, src, idx = row
-                    rw.writerow([int(start) + bytes_offset, int(end) + bytes_offset, id_, src, int(idx)])
-                    row_count += 1
-
-            bytes_offset += source_memmap.shape[0]
-            row_offset += row_count
+                    start_value = int(start)
+                    end_value = int(end)
+                    if path.selection_path is None:
+                        assert copy_start is not None
+                        output_start = copy_start + start_value
+                        output_end = copy_start + end_value
+                        input_processed_values = end_value
+                    else:
+                        output_start = token_offset
+                        output_end = token_offset + end_value - start_value
+                        target_memmap[output_start:output_end] = source_memmap[start_value:end_value]
+                        token_offset = output_end
+                        copied_values = end_value - start_value
+                        input_processed_values += copied_values
+                        token_copy_operations += 1
+                    input_documents += 1
+                    document_count += 1
+                    rw.writerow(
+                        [
+                            output_start,
+                            output_end,
+                            id_,
+                            src,
+                            int(idx),
+                        ]
+                    )
+                    if document_count % METADATA_ROW_LOG_INTERVAL == 0:
+                        now = time.monotonic()
+                        if now - last_progress_at >= PROGRESS_INTERVAL_SECONDS:
+                            publish_progress()
+                            last_progress_at = now
             del source_memmap
+            input_seconds = max(time.monotonic() - input_started_at, 1e-9)
+            finished_input_values = input_processed_values
+            completed_values += finished_input_values
+            input_processed_values = 0
+            publish_progress()
+            logger.info(
+                "merge %s input %s/%s done · %s · %s tokens · %s docs · " "%s tokens/s · %s docs/s · %s",
+                npy_destination.stem,
+                input_index,
+                len(paths),
+                view_type,
+                _human_count(finished_input_values),
+                _human_count(input_documents),
+                _human_count(round(finished_input_values / input_seconds)),
+                _human_count(round(input_documents / input_seconds)),
+                _elapsed(input_started_at),
+            )
+        flush_started_at = time.monotonic()
+        target_memmap.flush()
+        logger.info(
+            "merge %s flush · %s · %s",
+            npy_destination.stem,
+            _human_bytes(total_size),
+            _elapsed(flush_started_at),
+        )
+    if token_offset != total_values:
+        raise RuntimeError(
+            f"Reshard output length mismatch for {npy_destination}: "
+            f"wrote {token_offset}, expected {total_values}"
+        )
+    elapsed_seconds = max(time.monotonic() - started_at, 1e-9)
+    metadata_bytes = csv_destination.stat().st_size
+    result = MergeGroupResult(
+        destination=npy_destination,
+        output_bytes=total_size,
+        metadata_bytes=metadata_bytes,
+        document_count=document_count,
+        token_copy_operations=token_copy_operations,
+        elapsed_seconds=elapsed_seconds,
+    )
+    publish_progress(complete=True)
+    logger.info(
+        "merge %s done · %s tokens (%s) · %s metadata · %s docs · " "%s tokens/s · %s docs/s · %s",
+        npy_destination.stem,
+        _human_count(total_values),
+        _human_bytes(total_size),
+        _human_bytes(metadata_bytes),
+        _human_count(document_count),
+        _human_count(round(total_values / elapsed_seconds)),
+        _human_count(round(document_count / elapsed_seconds)),
+        _elapsed(started_at),
+    )
+    return result
 
 
 def group_paths_by_max_size(
     paths: list[TokensMetadataPaths],
     max_size_bytes: int,
 ) -> list[list[TokensMetadataPaths]]:
-    """
-    Group paths by max size.
+    """Group merge inputs into output shards of at most max_size_bytes each.
+
+    Every pass over the remaining inputs starts a new group and emits one copy of each
+    input, so repeated copies of an input are spread across groups.
     """
     counts: dict[TokensMetadataPaths, int] = {p: int(c) for p, c in Counter(paths).items()}
-    logger.info(
-        "Found %s unique paths from %s files; max repetition is %s",
-        len(counts),
-        len(paths),
-        max(counts.values()),
-    )
+    _log_merge_input_views(counts, len(paths))
 
     grouped_paths: list[list[TokensMetadataPaths]] = []
     while len(counts) > 0:
-        # add a fresh group
         grouped_paths.append([])
 
-        # partition in groups of max_num_files
         for path, _ in sorted(counts.items(), key=lambda x: -x[1]):
             if sum(p.size for p in grouped_paths[-1]) + path.size > max_size_bytes:
                 grouped_paths.append([path])
             else:
                 grouped_paths[-1].append(path)
 
-        # decrease counts, remove paths with 0 count.
         counts = {path: new_count for path, count in counts.items() if (new_count := count - 1) > 0}
 
     logger.info(
-        "By size: organized %s files into %s groups of max %.2f GB",
+        "Grouped %s merge input uses into %s output shards capped at %.2f GiB",
         len(paths),
         len(grouped_paths),
         max_size_bytes / 1024**3,
@@ -160,53 +394,55 @@ def group_paths_by_max_size(
     return grouped_paths
 
 
-def weighted_bucket_sample(values: list, count: int, weights: list[float]) -> list[int]:
-    """Sample bucket indices with optional weights."""
-
-    # Use the weighted sampling approach
-    keys = [random.random() * (w / sum(weights)) for w in weights]
-    indices = sorted(range(len(values)), key=lambda i: keys[i], reverse=True)[:count]
-    return indices
-
-
 def group_paths_by_max_num_files(
     paths: list[TokensMetadataPaths],
     max_num_files: int,
 ) -> list[list[TokensMetadataPaths]]:
-    """
-    Group paths by max number of files.
+    """Pack merge inputs into at most `max_num_files` size-balanced output shards.
+
+    Largest-first bin packing: every use of every input is expanded into a flat list,
+    shuffled so equal-sized views tie-break under the configured deterministic seed
+    rather than by dict order, sorted by descending size, and then placed in whichever
+    bucket is currently smallest, with ties broken randomly. The result is
+    `max_num_files` groups of roughly equal total bytes, minus any that stayed empty,
+    which happens when there are fewer inputs than buckets.
+
+    Repeated copies of the same input are not separated: when a manifest row has
+    `repeat_count > 1`, two or more copies of that shard can land in the same output
+    file.
     """
     counts = Counter(paths)
-    logger.info(
-        "Found %s unique paths from %s files; max repetition is %s",
-        len(counts),
-        len(paths),
-        max(counts.values()),
-    )
-
-    if (m := max(counts.values())) > max_num_files:
-        raise ValueError(f"One or more paths appear {m} times, exceeding max_num_files={max_num_files}")
+    _log_merge_input_views(counts, len(paths))
 
     grouped_paths: list[list[TokensMetadataPaths]] = [[] for _ in range(max_num_files)]
-    # Distribute each element across groups in round-robin fashion
-    for element, count in counts.items():
-        # sample count buckets out of max_num_files where we could put the element
-        # we sample with weights proportional to the number of elements in the bucket,
-        # so that we are more likely to sample buckets with fewer elements.
-        buckets = weighted_bucket_sample(
-            values=list(range(max_num_files)),
-            count=count,
-            weights=[1 / (len(grouped_paths[i]) + 1) for i in range(max_num_files)],
-        )
-        for bucket in buckets:
-            grouped_paths[bucket].append(element)
+    grouped_sizes = [0] * max_num_files
+    expanded_paths = [element for element, count in counts.items() for _ in range(count)]
+    random.shuffle(expanded_paths)
+    expanded_paths.sort(key=lambda path: path.size, reverse=True)
+    for element in expanded_paths:
+        minimum_size = min(grouped_sizes)
+        candidates = [index for index, size in enumerate(grouped_sizes) if size == minimum_size]
+        bucket = random.choice(candidates)
+        grouped_paths[bucket].append(element)
+        grouped_sizes[bucket] += element.size
 
-    # there is still a change that some buckets might be empty; we remove them.
     grouped_paths = [group for group in grouped_paths if len(group) > 0]
 
     return grouped_paths
 
 
+def _log_merge_input_views(
+    counts: Mapping[TokensMetadataPaths, int],
+    total_uses: int,
+) -> None:
+    logger.info(
+        "Merge inputs: %s uses · %s distinct inputs · max uses of any input: %s×",
+        total_uses,
+        len(counts),
+        max(counts.values()),
+    )
+
+
 def _get_worker_rank() -> int:
     """
     Returns an index for the current worker:
@@ -238,7 +474,7 @@ def _get_worker_rank() -> int:
 
 
 def _worker_init(seed: int):
-    """Initialize the random seed for the current worker."""
+    """Seed the global random module with `seed` offset by this worker's rank."""
     random.seed(seed + _get_worker_rank())
 
 
@@ -273,40 +509,102 @@ def merge_all_npys(
     else:
         raise ValueError("Either max_size_bytes or max_num_files must be provided")
 
+    total_bytes = sum(path.size for path in paths)
+    total_values = total_bytes // tokenizer.dtype.itemsize
     logger.info(
-        "Organizing %s files into %s groups using %s workers...", len(paths), len(grouped_paths), max_workers
+        "merge start · %s inputs · %s shards · %s tokens (%s) · %s workers",
+        len(paths),
+        len(grouped_paths),
+        _human_count(total_values),
+        _human_bytes(total_bytes),
+        max_workers,
     )
 
     init_fn = partial(_worker_init, seed=seed)
+    started_at = time.monotonic()
+    progress_lock = threading.Lock()
+    progress_by_shard: dict[str, MergeGroupProgress] = {}
+
+    def update_progress(update: MergeGroupProgress) -> None:
+        with progress_lock:
+            progress_by_shard[update.destination.stem] = update
 
     with ThreadPoolExecutor(max_workers=max_workers, initializer=init_fn) as pool:
-        futures = []
+        futures = set()
         for i, group in enumerate(grouped_paths):
             future = pool.submit(
                 merge_group,
                 paths=group,
                 destination=destination / f"{i:06d}.npy",
                 dtype=tokenizer.dtype,
+                progress=update_progress,
             )
-            futures.append(future)
-
-        for future in tqdm(as_completed(futures), total=len(futures), desc="Merging files"):
-            try:
-                future.result()
-            except Exception as e:
-                for future in futures:
-                    future.cancel()
-                raise e
-
-        logger.info("Done merging NumPy memmaps.")
+            futures.add(future)
+
+        pending = futures
+        completed_shards = 0
+        completed_metadata_bytes = 0
+        completed_documents = 0
+        while pending:
+            done, pending = wait(
+                pending,
+                timeout=PROGRESS_INTERVAL_SECONDS,
+                return_when=FIRST_COMPLETED,
+            )
+            for future in done:
+                try:
+                    result = future.result()
+                except Exception:
+                    for pending_future in pending:
+                        pending_future.cancel()
+                    raise
+                completed_shards += 1
+                completed_metadata_bytes += result.metadata_bytes
+                completed_documents += result.document_count
+            if pending:
+                elapsed_seconds = max(time.monotonic() - started_at, 1e-9)
+                with progress_lock:
+                    progress_snapshot = tuple(progress_by_shard.values())
+                processed_values = sum(update.processed_values for update in progress_snapshot)
+                processed_documents = sum(update.document_count for update in progress_snapshot)
+                active_shards = sum(not update.complete for update in progress_snapshot)
+                queued_shards = len(grouped_paths) - len(progress_snapshot)
+                percent = 100 * processed_values / total_values
+                queued = f" · {queued_shards} queued" if queued_shards else ""
+                logger.info(
+                    "merge %.1f%% · %s/%s tokens · %s tokens/s · %s docs/s · " "%s/%s done · %s active%s · %s",
+                    percent,
+                    _human_count(processed_values),
+                    _human_count(total_values),
+                    _human_count(round(processed_values / elapsed_seconds)),
+                    _human_count(round(processed_documents / elapsed_seconds)),
+                    completed_shards,
+                    len(grouped_paths),
+                    active_shards,
+                    queued,
+                    _elapsed(started_at),
+                )
+
+    elapsed_seconds = max(time.monotonic() - started_at, 1e-9)
+    logger.info(
+        "merge done · %s shards · %s tokens (%s) · %s metadata · %s docs · " "%s tokens/s · %s docs/s · %s",
+        len(grouped_paths),
+        _human_count(total_values),
+        _human_bytes(total_bytes),
+        _human_bytes(completed_metadata_bytes),
+        _human_count(completed_documents),
+        _human_count(round(total_values / elapsed_seconds)),
+        _human_count(round(completed_documents / elapsed_seconds)),
+        _elapsed(started_at),
+    )
 
 
 @dataclass
 class ReshardingPrefixConfig:
-    """
-    Configuration for a resharding source.
+    """A source prefix and the rate at which the npy files under it are sampled.
 
-    Can be used to download the files and compute file up/down sampling.
+    `download` copies an s3 prefix to a local directory; `take` lists the local
+    prefix and applies `sample_rate`.
     """
 
     prefix: str | Path
@@ -326,7 +624,7 @@ def download(self, local_prefix: str | Path) -> "ReshardingPrefixConfig":
             "s5cmd",
             "cp",
             "-sp",
-            "--if-source-newer",
+            "--no-clobber",
             f"{remote_prefix_no_star}/*",
             f"{local_prefix_no_trailing_slash}/",
         ]
@@ -360,18 +658,22 @@ def take(self) -> list[TokensMetadataPaths]:
 
         new_paths = []
 
-        # if the multiplier k is > 1, we first take ⌊k⌋ copies of each path.
+        # Take floor(sample_rate) copies of every path.
         if (repetition_rate := int(math.floor(self.sample_rate))) > 0:
             new_paths.extend(paths * repetition_rate)
 
-        # this is the remaining non-integer part of the sample rate; because the npys are actually uneven in
-        # size, the proper way to do this is to use an ILP solver; however, since usually most of the npys are
-        # of same size, we can just take a random sample.
+        # The fractional remainder is drawn as a uniform random sample of files. Exact
+        # token-level sampling would need an ILP solver, since the npys are uneven in
+        # size; this approximation holds when most npys are of similar size.
         if (residual_frac := self.sample_rate - repetition_rate) > 0:
             new_paths.extend(random.sample(paths, max(1, round(residual_frac * len(paths)))))
 
-        # sort by size
-        logger.info("Taking %s paths from %s using %s sample rate", len(new_paths), len(paths), self.sample_rate)
+        logger.info(
+            "Taking %s paths from %s using %s sample rate",
+            len(new_paths),
+            len(paths),
+            self.sample_rate,
+        )
         return new_paths
 
     def to_dict(self) -> dict:
@@ -382,46 +684,680 @@ def from_dict(cls, d: dict) -> "ReshardingPrefixConfig":
         return cls(**d)
 
 
-@dataclass
+@dataclass(frozen=True)
+class ReshardingManifestEntry:
+    npy_uri: str
+    metadata_uri: str
+    repeat_count: int
+    partial_target_uint32_values: int = 0
+    selection_seed: int = 0
+    selection_algorithm: str = DOCUMENT_SELECTION_ALGORITHM
+    npy_size_bytes: int | None = None
+    metadata_size_bytes: int | None = None
+    npy_etag: str = ""
+    metadata_etag: str = ""
+
+
+@dataclass(frozen=True)
+class PreparedManifestRow:
+    """One manifest row resolved to the local paths its later phases operate on.
+
+    `npy_path` and `metadata_path` are where the two objects live once
+    `ReshardingManifestConfig._plan_downloads` has run: inside the row's private
+    directory under the run prefix for remote rows, or wherever the manifest pointed
+    for local ones. `selection_path` is where a deterministic partial selection index
+    is written, and is unused by rows that do not request one. Expected object sizes
+    are not duplicated here; `entry` holds what the planner recorded.
+    """
+
+    entry: ReshardingManifestEntry
+    npy_path: Path
+    metadata_path: Path
+    selection_path: Path
+
+
+@dataclass(frozen=True)
+class DocumentSelectionJob:
+    manifest_index: int
+    source_uri: str
+    metadata_path: Path
+    selection_path: Path
+    source_uint32_values: int
+    target_uint32_values: int
+    seed: int
+
+
+@dataclass(frozen=True)
+class CompletedDocumentSelection:
+    manifest_index: int
+    result: DocumentSelectionResult
+
+
+def _run_document_selection(job: DocumentSelectionJob) -> CompletedDocumentSelection:
+    """Create one selection index in a process-pool worker."""
+
+    started_at = time.monotonic()
+    current_pass = ""
+    pass_started_at = started_at
+    source_document_count: int | None = None
+
+    def report_progress(pass_name: str, document_count: int, covered_values: int) -> None:
+        nonlocal current_pass, pass_started_at, source_document_count
+        now = time.monotonic()
+        if current_pass != pass_name:
+            current_pass = pass_name
+            pass_started_at = now
+            logger.info("Document selection %s %s started", job.source_uri, pass_name)
+        if document_count == 0:
+            return
+
+        pass_seconds = max(now - pass_started_at, 1e-9)
+        if pass_name == "pass 1/2" and covered_values == job.source_uint32_values:
+            source_document_count = document_count
+        if pass_name == "pass 2/2" and source_document_count is not None:
+            documents = f"{_human_count(document_count)}/{_human_count(source_document_count)} documents"
+        else:
+            documents = f"{_human_count(document_count)} documents"
+        logger.info(
+            "Document selection %s %s: %s · %s/%s values scanned · " "%s documents/s · %s values/s · elapsed %s",
+            job.source_uri,
+            pass_name,
+            documents,
+            _human_count(covered_values),
+            _human_count(job.source_uint32_values),
+            _human_count(round(document_count / pass_seconds)),
+            _human_count(round(covered_values / pass_seconds)),
+            _elapsed(pass_started_at),
+        )
+
+    result = create_document_selection(
+        metadata_path=job.metadata_path,
+        selection_path=job.selection_path,
+        source_uint32_values=job.source_uint32_values,
+        target_uint32_values=job.target_uint32_values,
+        seed=job.seed,
+        progress=report_progress,
+    )
+    elapsed_seconds = max(time.monotonic() - started_at, 1e-9)
+    scanned_documents = result.source_document_count * 2
+    scanned_values = job.source_uint32_values * 2
+    logger.info(
+        "Document selection %s complete: %s/%s documents selected · %s values selected "
+        "for %s requested (residual %+d) · %s documents/s · %s values/s · elapsed %s",
+        job.source_uri,
+        _human_count(result.selected_document_count),
+        _human_count(result.source_document_count),
+        _human_count(result.selected_uint32_values),
+        _human_count(result.requested_uint32_values),
+        result.target_residual_uint32_values,
+        _human_count(round(scanned_documents / elapsed_seconds)),
+        _human_count(round(scanned_values / elapsed_seconds)),
+        _elapsed(started_at),
+    )
+    return CompletedDocumentSelection(
+        manifest_index=job.manifest_index,
+        result=result,
+    )
+
+
+@dataclass(frozen=True)
+class ReshardingManifestConfig:
+    """An exact, locally stored manifest of token/metadata object pairs.
+
+    Version-one CSVs contain ``npy_uri``, ``metadata_uri``, and
+    ``repeat_count``. Version two may also request a deterministic partial copy
+    with ``partial_target_uint32_values`` and ``selection_seed``. Remote
+    objects are downloaded exactly once into a run-owned directory.
+    """
+
+    manifest: str | Path
+
+    def take(
+        self,
+        local_prefix: str | Path,
+        max_workers: int,
+        s5cmd_concurrency: int = DEFAULT_S5CMD_DOWNLOAD_CONCURRENCY,
+    ) -> list[TokensMetadataPaths]:
+        """Resolve this manifest into the merge inputs it names.
+
+        Six phases, in order: parse and validate the CSV, HEAD every remote object to
+        confirm it still matches the size and ETag the planner recorded, lay out one
+        local directory per row, run a single batched s5cmd download, re-verify the
+        downloaded sizes, and build any deterministic partial selections in a process
+        pool.
+
+        The returned list holds one entry per use of a source shard rather than one per
+        shard: a row with ``repeat_count`` 3 contributes three equal entries, and a row
+        that also requests a partial selection contributes one more.
+        """
+
+        manifest = Path(self.manifest)
+        if not manifest.is_file():
+            raise FileNotFoundError(f"Resharding manifest does not exist: {manifest}")
+
+        run_prefix = Path(local_prefix)
+        run_prefix.mkdir(parents=True, exist_ok=False)
+
+        rows = self._parse_manifest_rows(manifest)
+        self._verify_remote_sources(rows, max_workers=max_workers)
+        prepared, remote_commands = self._plan_downloads(
+            rows, local_prefix=run_prefix, s5cmd_concurrency=s5cmd_concurrency
+        )
+        self._download_remote_objects(prepared, remote_commands, local_prefix=run_prefix, max_workers=max_workers)
+        self._verify_downloaded_sizes(prepared)
+
+        row_paths, selection_jobs = self._build_row_views(prepared)
+        selection_views = self._run_selection_jobs(selection_jobs, prepared, max_workers=max_workers)
+        for manifest_index, selection_view in selection_views.items():
+            row_paths[manifest_index].append(selection_view)
+
+        paths = [path for per_row_paths in row_paths for path in per_row_paths]
+
+        if not paths:
+            raise RuntimeError(f"Manifest document selections produced no output: {manifest}")
+        selected_bytes = sum(path.size for path in paths)
+        logger.info(
+            "Manifest ready: %s merge input uses from %s source shards · %s planned output (%s uint32 values)",
+            len(paths),
+            len(rows),
+            _human_bytes(selected_bytes),
+            _human_count(selected_bytes // np.dtype(np.uint32).itemsize),
+        )
+        return paths
+
+    @classmethod
+    def _parse_manifest_rows(cls, manifest: Path) -> list[ReshardingManifestEntry]:
+        """Read and validate every row of the manifest CSV.
+
+        The header must carry at least the three version-one columns. Each row is then
+        validated by `_parse_manifest_row`, which names ``manifest:<line>`` in every
+        message. Line numbers start at 2 because line 1 is the header.
+        """
+
+        rows: list[ReshardingManifestEntry] = []
+        with manifest.open(newline="") as f:
+            reader = csv.DictReader(f)
+            expected = {"npy_uri", "metadata_uri", "repeat_count"}
+            if reader.fieldnames is None or not expected.issubset(reader.fieldnames):
+                raise ValueError(f"Manifest {manifest} must contain columns {sorted(expected)}")
+            for row_number, csv_row in enumerate(reader, start=2):
+                rows.append(cls._parse_manifest_row(csv_row, manifest=manifest, row_number=row_number))
+
+        if not rows:
+            raise ValueError(f"Resharding manifest is empty: {manifest}")
+
+        known_source_bytes = sum((entry.npy_size_bytes or 0) + (entry.metadata_size_bytes or 0) for entry in rows)
+        planned_uint32_values = sum(
+            (entry.npy_size_bytes or 0) // np.dtype(np.uint32).itemsize * entry.repeat_count
+            + entry.partial_target_uint32_values
+            for entry in rows
+        )
+        logger.info(
+            "Manifest loaded: %s source shards · %s objects · %s source data · "
+            "%s planned uint32 values · %s partial selections",
+            len(rows),
+            len(rows) * 2,
+            _human_bytes(known_source_bytes),
+            _human_count(planned_uint32_values),
+            sum(entry.partial_target_uint32_values > 0 for entry in rows),
+        )
+        return rows
+
+    @staticmethod
+    def _parse_manifest_row(
+        csv_row: Mapping[str, str],
+        manifest: Path,
+        row_number: int,
+    ) -> ReshardingManifestEntry:
+        """Validate one CSV row and turn it into a `ReshardingManifestEntry`.
+
+        Sizes and ETags are optional for local rows and mandatory for ``s3://`` ones,
+        which is what `_verify_remote_sources` and `_verify_downloaded_sizes` compare
+        against. A row requesting a partial selection also needs a uint32-aligned
+        source size strictly larger than the requested count, and must name a selection
+        algorithm this build implements.
+        """
+
+        npy_uri = csv_row["npy_uri"].strip()
+        metadata_uri = csv_row["metadata_uri"].strip()
+        try:
+            repeat_count = int(csv_row["repeat_count"])
+            partial_target = int(csv_row.get("partial_target_uint32_values") or 0)
+            selection_seed = int(csv_row.get("selection_seed") or 0)
+        except (TypeError, ValueError) as exc:
+            raise ValueError(f"Invalid sampling values on {manifest}:{row_number}") from exc
+        if not npy_uri.endswith(".npy") or not metadata_uri.endswith(".csv.gz"):
+            raise ValueError(f"Invalid object pair on {manifest}:{row_number}")
+        if any(ord(char) < 32 for char in npy_uri + metadata_uri):
+            raise ValueError(f"Control character in object URI on {manifest}:{row_number}")
+        if Path(npy_uri).stem != Path(Path(metadata_uri).stem).stem:
+            raise ValueError(f"Mismatched object pair on {manifest}:{row_number}")
+        if repeat_count < 0 or partial_target < 0:
+            raise ValueError(f"Sampling values cannot be negative on {manifest}:{row_number}")
+        if repeat_count == 0 and partial_target == 0:
+            raise ValueError(f"Manifest row selects no tokens on {manifest}:{row_number}")
+        try:
+            npy_size_bytes = int(csv_row["npy_size_bytes"]) if csv_row.get("npy_size_bytes") else None
+            metadata_size_bytes = (
+                int(csv_row["metadata_size_bytes"]) if csv_row.get("metadata_size_bytes") else None
+            )
+        except ValueError as exc:
+            raise ValueError(f"Invalid expected size on {manifest}:{row_number}") from exc
+        if npy_size_bytes is not None and npy_size_bytes <= 0:
+            raise ValueError(f"npy_size_bytes must be positive on {manifest}:{row_number}")
+        if metadata_size_bytes is not None and metadata_size_bytes <= 0:
+            raise ValueError(f"metadata_size_bytes must be positive on {manifest}:{row_number}")
+        if urlparse(npy_uri).scheme == "s3" and (npy_size_bytes is None or metadata_size_bytes is None):
+            raise ValueError(
+                f"Remote manifest rows require npy_size_bytes and metadata_size_bytes on "
+                f"{manifest}:{row_number}"
+            )
+        if partial_target:
+            if npy_size_bytes is None:
+                raise ValueError(f"Partial manifest rows require npy_size_bytes on {manifest}:{row_number}")
+            if npy_size_bytes % np.dtype(np.uint32).itemsize:
+                raise ValueError(f"Partial source size is not uint32-aligned on {manifest}:{row_number}")
+            source_values = npy_size_bytes // np.dtype(np.uint32).itemsize
+            if partial_target >= source_values:
+                raise ValueError(
+                    f"partial_target_uint32_values must be smaller than the source on {manifest}:{row_number}"
+                )
+            selection_algorithm = str(csv_row.get("selection_algorithm") or DOCUMENT_SELECTION_ALGORITHM)
+            if selection_algorithm != DOCUMENT_SELECTION_ALGORITHM:
+                raise ValueError(
+                    f"Unsupported selection_algorithm on {manifest}:{row_number}: {selection_algorithm}"
+                )
+        else:
+            selection_algorithm = DOCUMENT_SELECTION_ALGORITHM
+        return ReshardingManifestEntry(
+            npy_uri=npy_uri,
+            metadata_uri=metadata_uri,
+            repeat_count=repeat_count,
+            partial_target_uint32_values=partial_target,
+            selection_seed=selection_seed,
+            selection_algorithm=selection_algorithm,
+            npy_size_bytes=npy_size_bytes,
+            metadata_size_bytes=metadata_size_bytes,
+            npy_etag=str(csv_row.get("npy_etag", "")).strip('"'),
+            metadata_etag=str(csv_row.get("metadata_etag", "")).strip('"'),
+        )
+
+    @staticmethod
+    def _verify_remote_sources(rows: list[ReshardingManifestEntry], max_workers: int) -> None:
+        """HEAD every remote object and confirm its size and ETag still match.
+
+        A source object rewritten since the manifest was planned would change which
+        tokens this run emits, and for partial rows which documents get selected, so a
+        mismatch aborts before anything is downloaded. Rows whose recorded ETag is
+        empty are size-checked only. No-ops when the manifest is entirely local.
+        """
+
+        remote_expectations: list[tuple[str, int, str]] = []
+        for entry in rows:
+            if urlparse(entry.npy_uri).scheme == "s3":
+                assert entry.npy_size_bytes is not None
+                assert entry.metadata_size_bytes is not None
+                remote_expectations.extend(
+                    [
+                        (entry.npy_uri, entry.npy_size_bytes, entry.npy_etag),
+                        (entry.metadata_uri, entry.metadata_size_bytes, entry.metadata_etag),
+                    ]
+                )
+        if not remote_expectations:
+            return
+
+        client = boto3.client("s3")
+        validation_started_at = time.monotonic()
+        logger.info(
+            "Source validation started: checking size and ETag for %s objects",
+            len(remote_expectations),
+        )
+
+        def verify_remote(expectation: tuple[str, int, str]) -> None:
+            uri, expected_size, expected_etag = expectation
+            parsed = urlparse(uri)
+            response = client.head_object(Bucket=parsed.netloc, Key=parsed.path.lstrip("/"))
+            actual_size = int(response["ContentLength"])
+            actual_etag = str(response.get("ETag", "")).strip('"')
+            if actual_size != expected_size:
+                raise RuntimeError(
+                    f"Source object size changed before download: {uri}; "
+                    f"expected {expected_size}, found {actual_size}"
+                )
+            if expected_etag and actual_etag != expected_etag:
+                raise RuntimeError(
+                    f"Source object ETag changed before download: {uri}; "
+                    f"expected {expected_etag}, found {actual_etag}"
+                )
+
+        with ThreadPoolExecutor(max_workers=max_workers) as pool:
+            futures = [pool.submit(verify_remote, expectation) for expectation in remote_expectations]
+            checkpoint = max(1, len(futures) // 10)
+            for completed, future in enumerate(as_completed(futures), start=1):
+                future.result()
+                if completed == len(futures) or completed % checkpoint == 0:
+                    logger.info(
+                        "Source validation progress: %s/%s objects · elapsed %s",
+                        completed,
+                        len(futures),
+                        _elapsed(validation_started_at),
+                    )
+
+    @staticmethod
+    def _plan_downloads(
+        rows: list[ReshardingManifestEntry],
+        local_prefix: Path,
+        s5cmd_concurrency: int,
+    ) -> tuple[list[PreparedManifestRow], list[str]]:
+        """Give every row a private local directory and resolve its object paths.
+
+        Remote rows are copied into ``<local_prefix>/<index>/tokens.{npy,csv.gz}`` and
+        contribute two s5cmd `cp` command lines each. Local rows are read where they
+        already are and only checked for existence, so nothing is copied for them. A
+        single row may not mix schemes across its two objects.
+
+        Returns the prepared rows in manifest order, plus the s5cmd command lines,
+        which are empty when there is nothing to download.
+        """
+
+        prepared: list[PreparedManifestRow] = []
+        remote_commands: list[str] = []
+        for index, entry in enumerate(rows):
+            npy_uri = entry.npy_uri
+            metadata_uri = entry.metadata_uri
+            row_dir = local_prefix / f"{index:06d}"
+            row_dir.mkdir(exist_ok=False)
+            local_npy = row_dir / "tokens.npy"
+            local_metadata = row_dir / "tokens.csv.gz"
+
+            npy_scheme = urlparse(npy_uri).scheme
+            metadata_scheme = urlparse(metadata_uri).scheme
+            if npy_scheme == "s3" and metadata_scheme == "s3":
+                remote_commands.extend(
+                    [
+                        (
+                            "cp --show-progress --raw --no-clobber "
+                            f"--concurrency {s5cmd_concurrency} "
+                            f"{shlex.quote(npy_uri)} {shlex.quote(str(local_npy))}"
+                        ),
+                        (
+                            "cp --show-progress --raw --no-clobber "
+                            f"--concurrency {s5cmd_concurrency} "
+                            f"{shlex.quote(metadata_uri)} {shlex.quote(str(local_metadata))}"
+                        ),
+                    ]
+                )
+            elif npy_scheme in {"", "file"} and metadata_scheme in {"", "file"}:
+                local_npy = Path(urlparse(npy_uri).path if npy_scheme == "file" else npy_uri)
+                local_metadata = Path(urlparse(metadata_uri).path if metadata_scheme == "file" else metadata_uri)
+                if not local_npy.is_file() or not local_metadata.is_file():
+                    raise FileNotFoundError(f"Manifest object pair does not exist: {npy_uri}, {metadata_uri}")
+            else:
+                raise ValueError(f"Manifest row mixes unsupported URI schemes: {npy_uri}, {metadata_uri}")
+
+            prepared.append(
+                PreparedManifestRow(
+                    entry=entry,
+                    npy_path=local_npy,
+                    metadata_path=local_metadata,
+                    selection_path=row_dir / "selection.csv.gz",
+                )
+            )
+        return prepared, remote_commands
+
+    @staticmethod
+    def _download_remote_objects(
+        prepared: list[PreparedManifestRow],
+        remote_commands: list[str],
+        local_prefix: Path,
+        max_workers: int,
+    ) -> None:
+        """Run one batched s5cmd job and confirm it created every object.
+
+        All command lines go through a single command file, so one s5cmd process
+        handles the whole manifest. s5cmd can exit zero having skipped work, so the
+        object pairs are checked afterwards. No-ops when nothing needs downloading.
+        """
+
+        if not remote_commands:
+            return
+
+        commands_path = local_prefix / "s5cmd-download-commands.txt"
+        with commands_path.open("x") as f:
+            f.write("\n".join(remote_commands) + "\n")
+        cmd = ["s5cmd", "--stat", "--numworkers", str(max_workers), "run", str(commands_path)]
+        _run_s5cmd(cmd, "source download")
+
+        missing = [
+            (row.npy_path, row.metadata_path)
+            for row in prepared
+            if not row.npy_path.is_file() or not row.metadata_path.is_file()
+        ]
+        if missing:
+            raise RuntimeError(f"s5cmd completed without creating {len(missing)} manifest object pairs")
+
+    @staticmethod
+    def _verify_downloaded_sizes(prepared: list[PreparedManifestRow]) -> None:
+        """Re-check the on-disk sizes against the manifest after downloading.
+
+        `_verify_remote_sources` checks the source before the copy; this checks the
+        copy itself, so a truncated or partially resumed download cannot reach the
+        merge. Rows whose manifest recorded no size, only possible for local rows, are
+        skipped.
+        """
+
+        for row in prepared:
+            npy_path = row.npy_path
+            metadata_path = row.metadata_path
+            expected_npy_size = row.entry.npy_size_bytes
+            expected_metadata_size = row.entry.metadata_size_bytes
+            if expected_npy_size is not None and npy_path.stat().st_size != expected_npy_size:
+                raise RuntimeError(
+                    f"Downloaded NPY size does not match manifest: {npy_path}; "
+                    f"expected {expected_npy_size}, found {npy_path.stat().st_size}"
+                )
+            if expected_metadata_size is not None and metadata_path.stat().st_size != expected_metadata_size:
+                raise RuntimeError(
+                    f"Downloaded metadata size does not match manifest: {metadata_path}; "
+                    f"expected {expected_metadata_size}, found {metadata_path.stat().st_size}"
+                )
+
+    @staticmethod
+    def _build_row_views(
+        prepared: list[PreparedManifestRow],
+    ) -> tuple[list[list[TokensMetadataPaths]], list[DocumentSelectionJob]]:
+        """Expand each row's whole-file repeats and collect its selection job, if any.
+
+        Returns one list of merge inputs per manifest row, holding ``repeat_count``
+        copies of the full object, which may be zero for a purely partial row, plus one
+        `DocumentSelectionJob` for every row that also requests a partial selection.
+        The partial view is appended to its row later, once its selection index exists.
+        """
+
+        row_paths: list[list[TokensMetadataPaths]] = [[] for _ in prepared]
+        selection_jobs: list[DocumentSelectionJob] = []
+        for manifest_index, row in enumerate(prepared):
+            entry = row.entry
+            full_path = TokensMetadataPaths(str(row.npy_path), str(row.metadata_path))
+            row_paths[manifest_index].extend([full_path] * entry.repeat_count)
+            if entry.partial_target_uint32_values:
+                source_values = row.npy_path.stat().st_size // np.dtype(np.uint32).itemsize
+                selection_jobs.append(
+                    DocumentSelectionJob(
+                        manifest_index=manifest_index,
+                        source_uri=entry.npy_uri,
+                        metadata_path=row.metadata_path,
+                        selection_path=row.selection_path,
+                        source_uint32_values=source_values,
+                        target_uint32_values=entry.partial_target_uint32_values,
+                        seed=entry.selection_seed,
+                    )
+                )
+        return row_paths, selection_jobs
+
+    @staticmethod
+    def _run_selection_jobs(
+        selection_jobs: list[DocumentSelectionJob],
+        prepared: list[PreparedManifestRow],
+        max_workers: int,
+    ) -> dict[int, TokensMetadataPaths]:
+        """Build every requested partial selection index in a process pool.
+
+        Selection is CPU-bound, two passes over a shard's metadata, hence processes
+        rather than threads. Returns the partial merge-input view for each row that
+        selected at least one value, keyed by manifest index; a row whose selection
+        came back empty is omitted. Every row has at most one selection job, so a
+        manifest index appears at most once. No-ops when no row requested a selection.
+        """
+
+        if not selection_jobs:
+            return {}
+
+        selection_views: dict[int, TokensMetadataPaths] = {}
+        selection_workers = min(max_workers, len(selection_jobs))
+        selection_started_at = time.monotonic()
+        total_source_values = sum(job.source_uint32_values for job in selection_jobs)
+        total_target_values = sum(job.target_uint32_values for job in selection_jobs)
+        logger.info(
+            "Document selection started: %s source shards · %s processes · %s source values per pass · "
+            "%s requested values",
+            len(selection_jobs),
+            selection_workers,
+            _human_count(total_source_values),
+            _human_count(total_target_values),
+        )
+        completed_source_values = 0
+        completed_scanned_documents = 0
+        selected_values = 0
+        with ProcessPoolExecutor(max_workers=selection_workers) as pool:
+            futures = {pool.submit(_run_document_selection, job): job for job in selection_jobs}
+            for completed_count, future in enumerate(as_completed(futures), start=1):
+                try:
+                    completed = future.result()
+                except Exception:
+                    for pending_future in futures:
+                        pending_future.cancel()
+                    raise
+                job = futures[future]
+                selection = completed.result
+                if selection.selected_uint32_values:
+                    source_row = prepared[completed.manifest_index]
+                    selection_views[completed.manifest_index] = TokensMetadataPaths(
+                        str(source_row.npy_path),
+                        str(source_row.metadata_path),
+                        selection_path=str(selection.selection_path),
+                        selected_uint32_values=selection.selected_uint32_values,
+                    )
+                completed_source_values += job.source_uint32_values
+                completed_scanned_documents += selection.source_document_count * 2
+                selected_values += selection.selected_uint32_values
+                wall_seconds = max(time.monotonic() - selection_started_at, 1e-9)
+                logger.info(
+                    "Document selection progress: %s/%s source shards · %s/%s values selected · "
+                    "%s documents/s · %s values/s across both passes · elapsed %s",
+                    completed_count,
+                    len(selection_jobs),
+                    _human_count(selected_values),
+                    _human_count(total_target_values),
+                    _human_count(round(completed_scanned_documents / wall_seconds)),
+                    _human_count(round((completed_source_values * 2) / wall_seconds)),
+                    _elapsed(selection_started_at),
+                )
+        wall_seconds = max(time.monotonic() - selection_started_at, 1e-9)
+        logger.info(
+            "Document selection complete: %s source shards · %s/%s values selected · "
+            "%s documents/s · %s values/s across both passes · elapsed %s",
+            len(selection_jobs),
+            _human_count(selected_values),
+            _human_count(total_target_values),
+            _human_count(round(completed_scanned_documents / wall_seconds)),
+            _human_count(round((completed_source_values * 2) / wall_seconds)),
+            _elapsed(selection_started_at),
+        )
+        return selection_views
+
+    def to_dict(self) -> dict:
+        return {"manifest": str(self.manifest)}
+
+    @classmethod
+    def from_dict(cls, d: dict, base_dir: Path | None = None) -> "ReshardingManifestConfig":
+        path = Path(d["manifest"])
+        if base_dir is not None and not path.is_absolute():
+            path = base_dir / path
+        return cls(manifest=path)
+
+
+@dataclass(kw_only=True)
 class ReshardingConfig:
-    """Base configuration for resharding."""
+    """Base configuration for resharding.
+
+    Fields are keyword-only: `source_prefixes` is one of two alternative source lists
+    and is not the first field, so positional construction would bind arguments to the
+    wrong fields.
+    """
 
-    source_prefixes: list[ReshardingPrefixConfig]
     destination_prefix: str
+    source_prefixes: list[ReshardingPrefixConfig] = field(default_factory=list)
+    source_manifests: list[ReshardingManifestConfig] = field(default_factory=list)
     local_tempdir: str | Path | None = None
     max_size_bytes: int | None = None
     max_num_files: int | None = None
     max_workers: int = os.cpu_count() or 1
+    s5cmd_download_concurrency: int = DEFAULT_S5CMD_DOWNLOAD_CONCURRENCY
     random_seed: int = 42
     tokenizer_name_or_path: str = "allenai/dolma2-tokenizer"
+    # Tombstone: resharding always refuses a populated destination, and the field
+    # exists so that a config setting it True is rejected rather than ignored.
+    allow_existing_destination: bool = False
 
     def __post_init__(self):
         if self.max_size_bytes is not None and self.max_num_files is not None:
             raise ValueError("Cannot provide both max_size_bytes and max_num_files")
         if self.max_size_bytes is None and self.max_num_files is None:
             raise ValueError("Either max_size_bytes or max_num_files must be provided")
+        if not self.source_prefixes and not self.source_manifests:
+            raise ValueError("At least one source_prefix or source_manifest must be provided")
+        if self.allow_existing_destination:
+            raise ValueError("Overwriting an existing destination is not supported")
+        if self.max_workers <= 0:
+            raise ValueError("max_workers must be positive")
+        if self.s5cmd_download_concurrency <= 0:
+            raise ValueError("s5cmd_download_concurrency must be positive")
 
         if self.local_tempdir is None:
             logging.warning("No local tempdir provided; using a temporary directory")
-            self.local_tempdir = Path(mkdtemp())
         else:
             self.local_tempdir = Path(self.local_tempdir)
 
     def to_dict(self) -> dict:
         source_prefixes_dict = [p.to_dict() for p in self.source_prefixes]
-        return {**asdict(self), "source_prefixes": source_prefixes_dict}
+        source_manifests_dict = [m.to_dict() for m in self.source_manifests]
+        return {
+            **asdict(self),
+            "source_prefixes": source_prefixes_dict,
+            "source_manifests": source_manifests_dict,
+        }
 
     @classmethod
-    def from_dict(cls, d: dict) -> "ReshardingConfig":
+    def from_dict(cls, d: dict, base_dir: Path | None = None) -> "ReshardingConfig":
         source_prefixes = [ReshardingPrefixConfig.from_dict(p) for p in d.get("source_prefixes", [])]
+        source_manifests = [
+            ReshardingManifestConfig.from_dict(m, base_dir=base_dir) for m in d.get("source_manifests", [])
+        ]
         return cls(
-            source_prefixes=source_prefixes,
             destination_prefix=str(d["destination_prefix"]),
+            source_prefixes=source_prefixes,
+            source_manifests=source_manifests,
             local_tempdir=(Path(p) if (p := d.get("local_tempdir")) is not None else None),
             max_size_bytes=(int(s) if (s := d.get("max_size_bytes")) is not None else None),
             max_num_files=(int(n) if (n := d.get("max_num_files")) is not None else None),
             max_workers=int(d.get("max_workers", 1)),
+            s5cmd_download_concurrency=int(
+                d.get("s5cmd_download_concurrency", DEFAULT_S5CMD_DOWNLOAD_CONCURRENCY)
+            ),
             random_seed=int(d.get("random_seed", 42)),
+            tokenizer_name_or_path=str(d.get("tokenizer_name_or_path", "allenai/dolma2-tokenizer")),
+            allow_existing_destination=bool(d.get("allow_existing_destination", False)),
         )
 
     @classmethod
@@ -429,53 +1365,125 @@ def from_file(cls, file_path: str | Path) -> "ReshardingConfig":
         if file_path == "-":
             return cls.from_dict(yaml.safe_load(sys.stdin))
 
-        with open(file_path, "r") as f:
-            return cls.from_dict(yaml.safe_load(f))
+        path = Path(file_path)
+        with path.open("r") as f:
+            return cls.from_dict(yaml.safe_load(f), base_dir=path.parent)
 
 
 def upload_to_s3(local_prefix: str | Path, remote_prefix: str, max_workers: int):
-    """
-    Upload a local directory to S3.
+    """Upload a local directory to S3 with no-clobber semantics.
+
+    No-ops when `remote_prefix` is not an s3 URI.
     """
     if urlparse(remote_prefix).scheme != "s3":
         return
 
     local_prefix_no_star = re.sub(r"(/|/\*)$", "", str(local_prefix))
     remote_prefix_no_trailing_slash = str(remote_prefix).rstrip("/")
-    cmd = ["s5cmd", "cp", "-sp", f"{local_prefix_no_star}/*", f"{remote_prefix_no_trailing_slash}/"]
-    result = subprocess.run(cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    cmd = [
+        "s5cmd",
+        "--stat",
+        "--numworkers",
+        str(max_workers),
+        "cp",
+        "--no-clobber",
+        "--show-progress",
+        f"{local_prefix_no_star}/*",
+        f"{remote_prefix_no_trailing_slash}/",
+    ]
+    _run_s5cmd(cmd, "output upload")
+
+
+def destination_has_objects(destination: str | Path) -> bool:
+    """Return whether a destination already contains data.
+
+    S3 is queried with a read-only, one-key ``ListObjectsV2`` request. Local
+    destinations count as occupied when the path exists at all, including an empty
+    pre-created directory. Raises ValueError for an S3 bucket root or an unsupported
+    scheme.
+    """
 
-    if result.returncode != 0:
-        print(f"s5cmd failed with error: {result.stderr}")
-        raise Exception(f"Failed to upload files using s5cmd: {result.stderr}")
+    destination = str(destination)
+    parsed = urlparse(destination)
+    if parsed.scheme == "s3":
+        prefix = parsed.path.lstrip("/").rstrip("/")
+        if not parsed.netloc or not prefix:
+            raise ValueError("Refusing to materialize into an S3 bucket root")
+        response = boto3.client("s3").list_objects_v2(Bucket=parsed.netloc, Prefix=f"{prefix}/", MaxKeys=1)
+        return bool(response.get("Contents"))
+    if parsed.scheme not in {"", "file"}:
+        raise ValueError(f"Unsupported destination protocol: {parsed.scheme}")
+    path = Path(parsed.path if parsed.scheme == "file" else destination)
+    return os.path.lexists(path)
 
 
 def reshard(config: ReshardingConfig):
     random.seed(config.random_seed)
+    started_at = time.monotonic()
+    logger.info(
+        "Reshard started: %s source manifests · %s source prefixes · %s workers · destination=%s",
+        len(config.source_manifests),
+        len(config.source_prefixes),
+        config.max_workers,
+        config.destination_prefix,
+    )
+
+    if destination_has_objects(config.destination_prefix):
+        raise FileExistsError(f"Refusing to use existing destination: {config.destination_prefix}")
 
+    run_tempdir: Path | None = None
     try:
-        local_tempdir = Path(config.local_tempdir or mkdtemp())
-        local_tempdir.mkdir(parents=True, exist_ok=True)
+        if config.local_tempdir is None:
+            run_tempdir = Path(mkdtemp(prefix="dolma-reshard-"))
+        else:
+            temp_base = Path(config.local_tempdir)
+            temp_base.mkdir(parents=True, exist_ok=True)
+            run_tempdir = Path(mkdtemp(prefix="dolma-reshard-", dir=temp_base))
+
+        disk = shutil.disk_usage(run_tempdir)
+        logger.info(
+            "Working directory ready: %s · %s free of %s",
+            run_tempdir,
+            _human_bytes(disk.free),
+            _human_bytes(disk.total),
+        )
 
         local_output_dir = (
-            local_tempdir / "output"
+            run_tempdir / "output"
             if urlparse(config.destination_prefix).scheme == "s3"
             else Path(config.destination_prefix)
         )
 
-        # download the files
+        logger.info("Source preparation started")
         source_prefixes = [
-            source_prefix.download(local_tempdir / f"input/{i:06d}")
+            source_prefix.download(run_tempdir / f"prefix-input/{i:06d}")
             for i, source_prefix in enumerate(config.source_prefixes)
         ]
 
         # get repetition aware samples
         source_paths = [path for source_prefix in source_prefixes for path in source_prefix.take()]
+        for i, source_manifest in enumerate(config.source_manifests):
+            source_paths.extend(
+                source_manifest.take(
+                    run_tempdir / f"manifest-input/{i:06d}",
+                    max_workers=config.max_workers,
+                    s5cmd_concurrency=config.s5cmd_download_concurrency,
+                )
+            )
+
+        planned_bytes = sum(path.size for path in source_paths)
+        logger.info(
+            "Source preparation complete: %s merge input uses · %s distinct token/metadata "
+            "views · %s planned output (%s uint32 values) · elapsed %s",
+            len(source_paths),
+            len(set(source_paths)),
+            _human_bytes(planned_bytes),
+            _human_count(planned_bytes // np.dtype(np.uint32).itemsize),
+            _elapsed(started_at),
+        )
 
-        # make destination directory
-        local_output_dir.mkdir(parents=True, exist_ok=True)
+        local_output_dir.mkdir(parents=True, exist_ok=False)
 
-        # merge the files
         merge_all_npys(
             source_paths,
             destination=local_output_dir,
@@ -486,15 +1494,31 @@ def reshard(config: ReshardingConfig):
             seed=config.random_seed,
         )
 
-        # upload the files
+        output_files = [path for path in local_output_dir.rglob("*") if path.is_file()]
+        output_bytes = sum(path.stat().st_size for path in output_files)
+        logger.info(
+            "Local output ready: %s files · %s · elapsed %s",
+            len(output_files),
+            _human_bytes(output_bytes),
+            _elapsed(started_at),
+        )
+
         upload_to_s3(
             local_prefix=local_output_dir,
             remote_prefix=config.destination_prefix,
             max_workers=config.max_workers,
         )
+        logger.info(
+            "Reshard complete: %s files · %s · total elapsed %s",
+            len(output_files),
+            _human_bytes(output_bytes),
+            _elapsed(started_at),
+        )
 
     finally:
-        shutil.rmtree(local_tempdir)
+        if run_tempdir is not None:
+            logger.info("Removing run working directory: %s", run_tempdir)
+            shutil.rmtree(run_tempdir)
 
 
 def main():
diff --git a/scripts/dolma3p5_resharding/README.md b/scripts/dolma3p5_resharding/README.md
new file mode 100644
index 00000000..11d92b61
--- /dev/null
+++ b/scripts/dolma3p5_resharding/README.md
@@ -0,0 +1,49 @@
+# Dolma 3.5 resharding scripts
+
+Code layout. The run procedure is in `configs/dolma3p5-resharding/14t/README.md`.
+
+## Layout
+
+`workflow.py` holds the planning and validation logic. The other Python files are
+CLI wrappers around its entry points:
+
+| Script | Entry points | Writes |
+| --- | --- | --- |
+| `plan.py` | `plan_build`, `collect_inventory`, `propose_configs` | `01-plan/` |
+| `validate.py` | `validate_build` | nothing, checks the build |
+| `preflight.py` | `preflight_build` | `02-preflight/` |
+| `verify.py` | `verify_output` | `03-output-validation/` |
+| `materialize.py` | its own, calls `workflow.py` for selection and preflight | provisions workers, dispatches units |
+
+`output_report.py` renders the output-validation report.
+`setup_worker_storage.sh` runs on each worker to prepare local NVMe.
+
+## Scope
+
+- `python/dolma/tokenizer/reshard.py` and `document_selection.py` are the library
+  and contain no reference to Dolma 3.5 or the 14T target.
+- `scripts/resharding/dispatch.py` builds poormanray command lines and is
+  mixture-agnostic.
+- This directory is specific to the Dolma 3.5 14T campaign. It hard-codes the
+  default build path, cluster, and project, and reads the mix, catalog, and
+  settings from `configs/dolma3p5-resharding/14t/`.
+
+Target token count, source bucket, worker instance grid, output shard sizing, and
+residual bounds come from `settings.yaml`. The build-id prefix and default build
+path are in code.
+
+## Build directory
+
+`runs/dolma3p5-resharding/14t` by default, gitignored, marked by a `build.json`
+whose id derives from the mix and catalog hashes. Phases are `01-plan/` (stages
+`resolution`, `inventory`, `execution`), `02-preflight/`, and
+`03-output-validation/`. Tools refuse a directory that is not a recognized build
+of this workflow, and refuse a plan whose mix or catalog changed after creation.
+
+## Tests
+
+`tests/python/test_dolma3p5_resharding.py` covers this directory and the
+library's resharding safety behavior. `tests/python/test_resharding.py` covers
+`scripts/resharding/dispatch.py` and end-to-end resharding. Both import
+`scripts.*`, which requires `pythonpath = ["."]` in `pyproject.toml`. Neither
+needs network or AWS credentials.
diff --git a/scripts/dolma3p5_resharding/materialize.py b/scripts/dolma3p5_resharding/materialize.py
new file mode 100644
index 00000000..486c8019
--- /dev/null
+++ b/scripts/dolma3p5_resharding/materialize.py
@@ -0,0 +1,3133 @@
+# /// script
+# requires-python = ">=3.10"
+# dependencies = [
+#   "boto3",
+#   "poormanray",
+#   "PyYAML",
+#   "rich",
+# ]
+# ///
+
+"""Provision workers and materialize Dolma 3.5 execution units."""
+
+from __future__ import annotations
+
+import argparse
+import copy
+import csv
+import hashlib
+import json
+import os
+import re
+import shlex
+import shutil
+import subprocess
+import sys
+import time
+import uuid
+from collections import Counter, defaultdict, deque
+from collections.abc import Callable, Sequence
+from concurrent.futures import FIRST_COMPLETED, Future, ThreadPoolExecutor, as_completed, wait
+from contextlib import AbstractContextManager, nullcontext
+from dataclasses import dataclass, field
+from datetime import datetime, timezone
+from pathlib import Path
+from threading import Event, Lock
+from typing import Any, NoReturn, TypeVar
+from urllib.parse import urlparse
+
+import boto3
+from botocore.exceptions import BotoCoreError, ClientError
+from rich.console import Console
+from rich.table import Table
+from rich.text import Text
+
+scripts_root = Path(__file__).resolve().parents[1]
+if str(scripts_root) not in sys.path:
+    sys.path.insert(0, str(scripts_root))
+
+from resharding.dispatch import (
+    build_poormanray_create_command,
+    build_poormanray_instance_command,
+    build_poormanray_map_command,
+    build_poormanray_run_command,
+    build_poormanray_setup_dolma_command,
+    build_poormanray_transfer_command,
+)
+
+try:
+    from .workflow import (
+        DEFAULT_BUILD_PATH,
+        DEFAULT_REGION,
+        UINT32_BYTES,
+        PreparationError,
+        _count_label,
+        _filter_execution_units,
+        _human_byte_count,
+        _human_token_count,
+        _list_prefix,
+        _pair_metadata_key,
+        _unit_selection_digest,
+        _validate_execution_layout,
+        _validate_preparation_build,
+        normalize_region,
+        preflight_build,
+    )
+except ImportError:
+    from workflow import (
+        DEFAULT_BUILD_PATH,
+        DEFAULT_REGION,
+        UINT32_BYTES,
+        PreparationError,
+        _count_label,
+        _filter_execution_units,
+        _human_byte_count,
+        _human_token_count,
+        _list_prefix,
+        _pair_metadata_key,
+        _unit_selection_digest,
+        _validate_execution_layout,
+        _validate_preparation_build,
+        normalize_region,
+        preflight_build,
+    )
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description=__doc__,
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
+    )
+    parser.add_argument(
+        "--build",
+        type=Path,
+        default=DEFAULT_BUILD_PATH,
+        help="local preparation build directory",
+    )
+    selection = parser.add_mutually_exclusive_group(required=True)
+    selection.add_argument(
+        "--all",
+        action="store_true",
+        help="select every execution unit",
+    )
+    selection.add_argument(
+        "--category",
+        help="exact mix name, leaf ID, or MIX_NAME::CATEGORY_NAME selector",
+    )
+    selection.add_argument(
+        "--unit",
+        help="select one exact execution-unit ID",
+    )
+    selection.add_argument(
+        "--list-categories",
+        nargs="?",
+        const="",
+        metavar="FILTER",
+        help="list category selectors, optionally filtered by a case-insensitive substring",
+    )
+    parser.add_argument(
+        "--exclude-category",
+        action="append",
+        default=[],
+        metavar="SELECTOR",
+        help=(
+            "exclude an exact mix name, leaf ID, or MIX_NAME::CATEGORY_NAME from --all; "
+            "repeat for multiple completed categories"
+        ),
+    )
+    parser.add_argument(
+        "--cluster",
+        default="dolma3p5-14t",
+        help="poormanray cluster name",
+    )
+    parser.add_argument(
+        "--region",
+        default=os.environ.get("PMR_REGION") or DEFAULT_REGION,
+        help="poormanray region; PMR_REGION overrides the us-east-1 default",
+    )
+    parser.add_argument("--profile", help="AWS profile used for the worker lifecycle")
+    parser.add_argument(
+        "--project",
+        default="oe-other",
+        help="worker project tag",
+    )
+    parser.add_argument(
+        "-j",
+        "--parallelism",
+        type=lambda value: _positive_integer(value, "parallelism"),
+        default=128,
+        help="maximum number of materialization workers kept active concurrently",
+    )
+    parser.add_argument(
+        "--provision-batch-size",
+        type=lambda value: _positive_integer(value, "provision-batch-size"),
+        default=DEFAULT_PROVISION_BATCH_SIZE,
+        help="maximum VM lifecycle requests submitted concurrently in one batch",
+    )
+    parser.add_argument(
+        "--provision-batch-delay-seconds",
+        type=lambda value: _nonnegative_float(value, "provision-batch-delay-seconds"),
+        default=DEFAULT_PROVISION_BATCH_DELAY_SECONDS,
+        metavar="SECONDS",
+        help="delay between VM lifecycle batches so provider API quotas can refill",
+    )
+    parser.add_argument(
+        "--bootstrap-parallelism",
+        type=lambda value: _positive_integer(value, "bootstrap-parallelism"),
+        default=32,
+        help="maximum ready workers bootstrapped and dispatched concurrently",
+    )
+    parser.add_argument(
+        "--readiness-poll-seconds",
+        type=lambda value: _positive_integer(value, "readiness-poll-seconds"),
+        default=10,
+        metavar="SECONDS",
+        help="interval for detecting newly ready workers",
+    )
+    parser.add_argument(
+        "--instance-type",
+        help="override the worker type selected by the execution plan",
+    )
+    parser.add_argument(
+        "--root-storage-type",
+        default="gp3",
+        help="worker root-volume type",
+    )
+    parser.add_argument(
+        "--root-storage-size",
+        type=lambda value: _positive_integer(value, "root-storage-size"),
+        default=200,
+        metavar="GIB",
+        help="worker root-volume size in GiB",
+    )
+    parser.add_argument(
+        "--storage-layout",
+        choices=("auto", "single", "raid0"),
+        default="auto",
+        help="local NVMe layout; auto uses the execution plan",
+    )
+    parser.add_argument(
+        "--ssh-key-path",
+        type=Path,
+        help="SSH private key passed to poormanray; its normal default is used when omitted",
+    )
+    parser.add_argument(
+        "--completion-poll-seconds",
+        type=lambda value: _positive_integer(value, "completion-poll-seconds"),
+        default=30,
+        metavar="SECONDS",
+        help="interval for checking whether materialization workers have stopped",
+    )
+    parser.add_argument(
+        "-v",
+        "--verbose",
+        action="store_true",
+        help="stream poormanray output and periodically show worker resharding logs",
+    )
+    parser.add_argument(
+        "--preflight",
+        action="store_true",
+        help="run the selected read-only preflight immediately before --execute",
+    )
+    mode = parser.add_mutually_exclusive_group()
+    mode.add_argument(
+        "--dry-run",
+        action="store_true",
+        help="stage and print the complete worker lifecycle without invoking poormanray",
+    )
+    mode.add_argument(
+        "--execute",
+        action="store_true",
+        help="provision, prepare, and dispatch poormanray workers after a passing preflight",
+    )
+    return parser
+
+
+def _positive_integer(value: str, name: str) -> int:
+    try:
+        parsed = int(value)
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError(f"{name} must be an integer") from exc
+    if parsed <= 0:
+        raise argparse.ArgumentTypeError(f"{name} must be positive")
+    return parsed
+
+
+def _nonnegative_float(value: str, name: str) -> float:
+    try:
+        parsed = float(value)
+    except ValueError as exc:
+        raise argparse.ArgumentTypeError(f"{name} must be a number") from exc
+    if parsed < 0:
+        raise argparse.ArgumentTypeError(f"{name} must not be negative")
+    return parsed
+
+
+REPOSITORY_ROOT = Path(__file__).resolve().parents[2]
+WORKER_STORAGE_SCRIPT = Path(__file__).with_name("setup_worker_storage.sh")
+RESHARD_MODULE = REPOSITORY_ROOT / "python/dolma/tokenizer/reshard.py"
+DOCUMENT_SELECTION_MODULE = (
+    REPOSITORY_ROOT / "python/dolma/tokenizer/document_selection.py"
+)
+REMOTE_STORAGE_SCRIPT = "/tmp/dolma3p5-setup-worker-storage.sh"
+REMOTE_RESHARD_MODULE = "/tmp/dolma3p5-runtime/reshard.py"
+REMOTE_DOCUMENT_SELECTION_MODULE = "/tmp/dolma3p5-runtime/document_selection.py"
+# setup_worker_storage.sh mounts the worker's local NVMe here (DOLMA_WORK_MOUNT).
+WORKER_LOCAL_MOUNT = "/mnt/dolma"
+# Must equal the `plan.py --local-temp-root` used to build the plan. That value
+# is recorded only in each per-unit config YAML (as `local_tempdir`), not in
+# config-index.csv or dataset-layout.json, so a mismatch is first detected at
+# worker bootstrap.
+WORKER_LOCAL_TEMP_ROOT = f"{WORKER_LOCAL_MOUNT}/dolma3p5-resharding"
+# Interpreter installed on every worker by poormanray's setup-dolma step.
+WORKER_PYTHON_BIN = "$HOME/.venv/bin/python"
+# Per-run status and log directory. The launcher payload writes it and the
+# progress monitor reads it, so both must resolve to the same remote path.
+WORKER_STATUS_ROOT = "$HOME/dolma3p5-resharding-status"
+ANSI_ESCAPE = re.compile(r"\x1b\[[0-?]*[ -/]*[@-~]")
+DOLMA_LOG_PREFIX = re.compile(
+    r"^\[(?P<timestamp>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}) [^\]]+ "
+    r"(?P<level>TRACE|DEBUG|INFO|WARNING|ERROR|CRITICAL)\]\s*"
+)
+COMPACT_LOG_PREFIX = re.compile(
+    r"^\[(?P<timestamp>\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2})\]\s*"
+)
+WAIT_STATUS = re.compile(
+    r"Waiting for instances\.\.\.\s*(?P<ready>\d+/\d+ ready)"
+    r"(?:\s*\((?P<elapsed>[^)]+)\))?",
+    re.IGNORECASE,
+)
+AWS_ACCESS_KEY = re.compile(r"\b(?:AKIA|ASIA)[A-Z0-9]{16}\b")
+SECRET_VALUE = re.compile(
+    r"(?i)\b(aws_secret_access_key|aws_session_token|secret_access_key)"
+    r"(\s*[=:]\s*)\S+"
+)
+PROCESS_TAIL_LINES = 12
+WORKER_LOG_PAGE_BYTES = 16 * 1024
+# EC2 accepts at most this many instance IDs per DescribeInstanceStatus call.
+DESCRIBE_INSTANCE_STATUS_ID_LIMIT = 100
+# Poormanray resolves a discovery group to instance IDs before applying explicit
+# ones, so a group must stay under the limit above, with a 10-instance margin.
+PMR_DISCOVERY_WORKER_LIMIT = DESCRIBE_INSTANCE_STATUS_ID_LIMIT - 10
+LIFECYCLE_MAX_ATTEMPTS = 5
+# Argparse defaults, also used as getattr fallbacks for hand-built Namespaces.
+DEFAULT_PROVISION_BATCH_SIZE = 5
+DEFAULT_PROVISION_BATCH_DELAY_SECONDS = 3.0
+# EC2 applies tags asynchronously; wait this long for them to become readable.
+TAG_PROPAGATION_TIMEOUT_SECONDS = 60
+# One status/log page must come back before the next monitor poll is due.
+WORKER_LOG_COMMAND_TIMEOUT_SECONDS = 120
+# Instance IDs share a long common prefix, so only the tail identifies a worker.
+WORKER_TAG_LENGTH = 6
+WORKER_LOG_STYLES = (
+    "bold bright_cyan",
+    "bold bright_magenta",
+    "bold bright_green",
+    "bold bright_yellow",
+    "bold bright_blue",
+    "bold bright_red",
+)
+
+
+@dataclass(frozen=True)
+class ClusterInstance:
+    instance_id: str
+    state: str
+    instance_type: str
+    project: str | None
+    name: str = ""
+
+
+@dataclass(frozen=True)
+class MaterializedUnitCheck:
+    unit_id: str
+    actual_uint32_values: int
+    npy_count: int
+    metadata_count: int
+    problems: tuple[str, ...]
+
+
+@dataclass(frozen=True)
+class PlannedWorkerGroup:
+    instance_type: str
+    storage_layout: str
+    rows: tuple[dict[str, str], ...]
+
+
+@dataclass(frozen=True)
+class MaterializationGroup:
+    args: argparse.Namespace
+    rows: tuple[dict[str, str], ...]
+    script_dir: Path
+    worker_count: int
+
+
+@dataclass(frozen=True)
+class WorkerAssignment:
+    group: MaterializationGroup
+    rows: tuple[dict[str, str], ...]
+    script_dir: Path
+
+
+_BatchItem = TypeVar("_BatchItem")
+
+
+def _provision_batch_size(args: argparse.Namespace) -> int:
+    """Return the lifecycle batch size, defaulted for hand-built Namespaces."""
+
+    return int(getattr(args, "provision_batch_size", DEFAULT_PROVISION_BATCH_SIZE))
+
+
+def _provision_batch_delay(args: argparse.Namespace) -> float:
+    """Return the seconds between lifecycle batches, defaulted for hand-built Namespaces."""
+
+    return float(
+        getattr(
+            args,
+            "provision_batch_delay_seconds",
+            DEFAULT_PROVISION_BATCH_DELAY_SECONDS,
+        )
+    )
+
+
+def _lifecycle_batches(
+    items: Sequence[_BatchItem], batch_size: int
+) -> list[list[_BatchItem]]:
+    """Split items into batches of at most batch_size."""
+
+    if batch_size <= 0:
+        raise ValueError("batch_size must be positive")
+    return [
+        list(items[offset : offset + batch_size])
+        for offset in range(0, len(items), batch_size)
+    ]
+
+
+def _worker_tag(instance_id: str) -> str:
+    """Return the instance-ID suffix used as a worker's display tag."""
+
+    return instance_id[-WORKER_TAG_LENGTH:]
+
+
+def _remote_status_root(status_run_id: str) -> str:
+    """Return the remote status directory shared by the launchers and the monitor."""
+
+    return f"{WORKER_STATUS_ROOT}/{status_run_id}"
+
+
+def _worker_log_line(
+    tag: str,
+    style: str,
+    message: str,
+    *,
+    timestamp: str | None = None,
+    bold: bool = False,
+) -> Text:
+    line = Text()
+    rendered_timestamp = timestamp or datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M:%S")
+    line.append(f"[{rendered_timestamp}]", style="dim")
+    line.append(" ")
+    line.append(f"[{tag}]", style=style)
+    line.append(" ")
+    line.append(message, style="bold" if bold else None)
+    return line
+
+
+def _worker_log_parts(message: str) -> tuple[str | None, str]:
+    """Return one worker line's source timestamp and its text without logger metadata."""
+
+    match = DOLMA_LOG_PREFIX.match(message)
+    if match is not None:
+        rendered = message[match.end() :]
+        if match.group("level") in {"WARNING", "ERROR", "CRITICAL"}:
+            rendered = f"{match.group('level').lower()} · {rendered}"
+        return match.group("timestamp"), rendered
+
+    compact_match = COMPACT_LOG_PREFIX.match(message)
+    if compact_match is not None:
+        return compact_match.group("timestamp"), message[compact_match.end() :]
+    return None, message
+
+
+def _worker_log_message(message: str) -> str:
+    """Remove Dolma's module and level metadata from one worker line."""
+
+    return _worker_log_parts(message)[1]
+
+
+def _describe_cluster_instances(
+    cluster: str, region: str, profile: str | None = None
+) -> list[ClusterInstance]:
+    """Return the instances tagged cluster, project, or Project with this cluster name."""
+
+    session = boto3.Session(profile_name=profile, region_name=region)
+    client = session.client("ec2", region_name=region)
+    descriptions: dict[str, dict[str, Any]] = {}
+    states = ["pending", "running", "stopping", "stopped"]
+    for tag_name in ("cluster", "project", "Project"):
+        paginator = client.get_paginator("describe_instances")
+        for page in paginator.paginate(
+            Filters=[
+                {"Name": "instance-state-name", "Values": states},
+                {"Name": f"tag:{tag_name}", "Values": [cluster]},
+            ]
+        ):
+            for reservation in page.get("Reservations", []):
+                for description in reservation.get("Instances", []):
+                    instance_id = description.get("InstanceId")
+                    if isinstance(instance_id, str):
+                        descriptions[instance_id] = description
+
+    instances = []
+    for instance_id, description in descriptions.items():
+        tags = {
+            str(tag.get("Key")): str(tag.get("Value"))
+            for tag in description.get("Tags", [])
+            if tag.get("Key") is not None and tag.get("Value") is not None
+        }
+        instances.append(
+            ClusterInstance(
+                instance_id=instance_id,
+                state=str(description.get("State", {}).get("Name", "unknown")),
+                instance_type=str(description.get("InstanceType", "unknown")),
+                project=tags.get("ai2-project") or tags.get("project"),
+                name=tags.get("Name", ""),
+            )
+        )
+    return sorted(instances, key=lambda instance: instance.instance_id)
+
+
+def _retag_cluster_instances(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    *,
+    names: dict[str, str] | None = None,
+) -> None:
+    """Apply the discovery, accounting, and cluster tags to each worker.
+
+    Blocks until the provider reports every tag and raises PreparationError
+    after TAG_PROPAGATION_TIMEOUT_SECONDS.
+    """
+
+    if not instance_ids:
+        return
+    session = boto3.Session(profile_name=args.profile, region_name=args.region)
+    client = session.client("ec2", region_name=args.region)
+    required_tags = {
+        # Poormanray discovers instances through this tag before applying
+        # explicit instance IDs; each discovery group must stay below the
+        # 100-ID DescribeInstanceStatus limit.
+        "project": _pmr_discovery_name(args),
+        "ai2-project": args.project,
+        "cluster": args.cluster,
+    }
+    client.create_tags(
+        Resources=list(instance_ids),
+        Tags=[{"Key": key, "Value": value} for key, value in required_tags.items()],
+    )
+    for instance_id, name in sorted((names or {}).items()):
+        client.create_tags(
+            Resources=[instance_id],
+            Tags=[{"Key": "Name", "Value": name}],
+        )
+
+    deadline = time.monotonic() + TAG_PROPAGATION_TIMEOUT_SECONDS
+    while True:
+        response = client.describe_instances(InstanceIds=list(instance_ids))
+        observed: dict[str, dict[str, str]] = {}
+        for reservation in response.get("Reservations", []):
+            for description in reservation.get("Instances", []):
+                instance_id = str(description.get("InstanceId", ""))
+                observed[instance_id] = {
+                    str(tag.get("Key")): str(tag.get("Value"))
+                    for tag in description.get("Tags", [])
+                    if tag.get("Key") is not None and tag.get("Value") is not None
+                }
+        if all(
+            all(
+                observed.get(instance_id, {}).get(key) == value
+                for key, value in required_tags.items()
+            )
+            and (
+                names is None
+                or instance_id not in names
+                or observed[instance_id].get("Name") == names[instance_id]
+            )
+            for instance_id in instance_ids
+        ):
+            return
+        if time.monotonic() >= deadline:
+            raise PreparationError(
+                "Timed out waiting for required worker tags on "
+                + ", ".join(sorted(instance_ids))
+            )
+        time.sleep(2)
+
+
+def _clean_process_line(raw_line: str) -> str:
+    line = ANSI_ESCAPE.sub("", raw_line).replace("\r", "").strip()
+    line = AWS_ACCESS_KEY.sub("<redacted-aws-key>", line)
+    return SECRET_VALUE.sub(r"\1\2<redacted>", line)
+
+
+def _status_detail(stage: str, line: str) -> str | None:
+    if not line:
+        return None
+    if stage == "wait for workers":
+        match = WAIT_STATUS.search(line)
+        if match:
+            elapsed = match.group("elapsed")
+            return (
+                f"{match.group('ready')} · {elapsed}"
+                if elapsed
+                else match.group("ready")
+            )
+        if line.startswith(("·", "•")):
+            return None
+    if line.startswith("[INFO]"):
+        return None
+    if line.startswith(("Instance ", "stdout:", "stderr:")):
+        return None
+    return line if len(line) <= 120 else f"{line[:117]}..."
+
+
+def _stage_status(stage: str, detail: str | None = None) -> Text:
+    status = Text(stage, style="bold")
+    if detail:
+        status.append("  ", style="dim")
+        status.append(detail, style="dim")
+    return status
+
+
+def _elapsed_time(started_at: float) -> str:
+    elapsed = max(0, round(time.monotonic() - started_at))
+    minutes, seconds = divmod(elapsed, 60)
+    if minutes:
+        return f"{minutes}m {seconds:02d}s"
+    return f"{seconds}s"
+
+
+def _run_compact_process(
+    stage: str,
+    command: Sequence[str],
+    *,
+    console: Console | None = None,
+    verbose: bool = False,
+    live: bool = True,
+) -> int:
+    """Run a command with an in-place status line and return its exit code.
+
+    On a non-zero exit the last PROCESS_TAIL_LINES lines of output are printed.
+    """
+
+    output = console or Console(stderr=True, highlight=False)
+    started_at = time.monotonic()
+    tail: deque[str] = deque(maxlen=PROCESS_TAIL_LINES)
+    process: subprocess.Popen[str] | None = None
+    live_status = (
+        output.status(_stage_status(stage), spinner="dots")
+        if output.is_terminal and not verbose and live
+        else None
+    )
+    if verbose:
+        output.print(Text.assemble(("→", "cyan"), " ", (stage, "bold")))
+    elif live_status is None:
+        output.print(Text.assemble(("…", "cyan"), " ", (stage, "bold")))
+
+    try:
+        process = subprocess.Popen(
+            command,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            bufsize=1,
+        )
+        if live_status is not None:
+            live_status.start()
+        assert process.stdout is not None
+        for raw_line in process.stdout:
+            line = _clean_process_line(raw_line)
+            if not line:
+                continue
+            if not tail or tail[-1] != line:
+                tail.append(line)
+            if verbose:
+                output.print(Text(f"  {line}", style="dim"))
+            detail = _status_detail(stage, line)
+            if live_status is not None and detail:
+                live_status.update(_stage_status(stage, detail))
+        return_code = process.wait()
+    except BaseException:
+        if process is not None and process.poll() is None:
+            process.terminate()
+            try:
+                process.wait(timeout=5)
+            except subprocess.TimeoutExpired:
+                process.kill()
+                process.wait()
+        raise
+    finally:
+        if live_status is not None:
+            live_status.stop()
+
+    duration = _elapsed_time(started_at)
+    if return_code == 0:
+        output.print(
+            Text.assemble(("✓", "bold green"), " ", stage, (f"  {duration}", "dim"))
+        )
+        return return_code
+
+    output.print(Text.assemble(("✗", "bold red"), " ", stage, (f"  {duration}", "dim")))
+    if tail:
+        output.print(Text("last output:", style="bold red"))
+        for line in tail:
+            output.print(Text(f"  {line[:240]}", style="dim"))
+    return return_code
+
+
+def _run_lifecycle_command(
+    stage: str,
+    command: Sequence[str],
+    *,
+    console: Console | None = None,
+    verbose: bool = False,
+    live: bool = True,
+) -> None:
+    try:
+        return_code = _run_compact_process(
+            stage,
+            command,
+            console=console,
+            verbose=verbose,
+            live=live,
+        )
+    except OSError as exc:
+        raise PreparationError(f"could not start {stage}: {exc}") from exc
+    if return_code:
+        raise PreparationError(f"{stage} failed with exit code {return_code}")
+
+
+def _instance_options(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> dict[str, Any]:
+    return {
+        # Poormanray selects instances through the `project` tag, passed as
+        # --name. `ai2-project` is the accounting project, passed as --project.
+        "cluster": _pmr_discovery_name(args),
+        "project": args.project,
+        "region": args.region,
+        "instance_ids": instance_ids,
+        "parallelism": (
+            min(args.parallelism, len(instance_ids))
+            if instance_ids
+            else args.parallelism
+        ),
+        "ssh_key_path": args.ssh_key_path,
+    }
+
+
+def _pmr_discovery_name(args: argparse.Namespace) -> str:
+    """Return the poormanray discovery group name for one worker class.
+
+    Falls back to the plain cluster name when the instance type or the storage
+    layout is not pinned.
+    """
+
+    instance_type = getattr(args, "instance_type", None)
+    storage_layout = getattr(args, "storage_layout", None)
+    if not instance_type or not storage_layout or storage_layout == "auto":
+        return args.cluster
+    suffix = re.sub(
+        r"[^a-z0-9]+",
+        "-",
+        f"{instance_type}-{storage_layout}".lower(),
+    ).strip("-")
+    return f"{args.cluster}-{suffix}"
+
+
+def _create_command(
+    args: argparse.Namespace,
+    number: int,
+    *,
+    detach: bool = False,
+) -> list[str]:
+    return build_poormanray_create_command(
+        cluster=args.cluster,
+        project=args.project,
+        region=args.region,
+        number=number,
+        instance_type=args.instance_type,
+        storage_type=args.root_storage_type,
+        storage_size_gib=args.root_storage_size,
+        parallelism=min(_provision_batch_size(args), number),
+        detach=detach,
+        ssh_key_path=args.ssh_key_path,
+    )
+
+
+def _provision_batches(worker_count: int, batch_size: int) -> tuple[int, ...]:
+    """Return the worker count for each creation batch."""
+
+    if worker_count < 0:
+        raise ValueError("worker_count must not be negative")
+    return tuple(
+        len(batch) for batch in _lifecycle_batches(range(worker_count), batch_size)
+    )
+
+
+def _wait_command(args: argparse.Namespace, instance_ids: Sequence[str]) -> list[str]:
+    return build_poormanray_instance_command(
+        "wait",
+        cluster=_pmr_discovery_name(args),
+        project=args.project,
+        region=args.region,
+        instance_ids=instance_ids,
+        ssh_key_path=args.ssh_key_path,
+    )
+
+
+def _resume_command(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    *,
+    detach: bool = False,
+) -> list[str]:
+    options = _instance_options(args, instance_ids)
+    options.pop("ssh_key_path")
+    return build_poormanray_instance_command("resume", detach=detach, **options)
+
+
+def _resume_workers_in_batches(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    *,
+    detach: bool = True,
+    console: Console | None = None,
+) -> None:
+    """Resume stopped workers in batches of --provision-batch-size.
+
+    Sleeps --provision-batch-delay-seconds between batches.
+    """
+
+    ordered = sorted(set(instance_ids))
+    if not ordered:
+        return
+    delay = _provision_batch_delay(args)
+    batches = _lifecycle_batches(ordered, _provision_batch_size(args))
+    output = console or Console(stderr=True, highlight=False)
+    discovery_name = _pmr_discovery_name(args)
+
+    for batch_index, batch in enumerate(batches, start=1):
+        _resume_worker_batch(
+            args,
+            batch,
+            stage=(
+                f"resume {discovery_name} batch "
+                f"{batch_index:,}/{len(batches):,}"
+            ),
+            detach=detach,
+            console=output,
+        )
+
+        if batch_index < len(batches) and delay:
+            time.sleep(delay)
+
+
+def _worker_states(args: argparse.Namespace) -> dict[str, str]:
+    """Return the provider's current state for every worker in the cluster."""
+
+    return {
+        instance.instance_id: instance.state
+        for instance in _describe_cluster_instances(
+            args.cluster, args.region, args.profile
+        )
+    }
+
+
+def _retry_lifecycle_batch(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    *,
+    run_attempt: Callable[[list[str]], bool],
+    describe_states: Callable[[list[str]], dict[str, str]],
+    retry_states: frozenset[str],
+    on_exhausted: Callable[[list[str], int], None],
+    on_retry: Callable[[list[str], float], None] | None = None,
+) -> None:
+    """Retry one lifecycle batch until no worker still needs the change.
+
+    `run_attempt` returns True when the batch succeeded. After a failed attempt
+    only the workers the provider reports in `retry_states` are retried, with
+    the inter-batch delay doubling per attempt. `on_exhausted` runs in place of
+    the attempt after LIFECYCLE_MAX_ATTEMPTS.
+    """
+
+    remaining = list(instance_ids)
+    delay = _provision_batch_delay(args)
+    for attempt in range(1, LIFECYCLE_MAX_ATTEMPTS + 1):
+        if run_attempt(remaining):
+            return
+        states = describe_states(remaining)
+        remaining = [
+            instance_id
+            for instance_id in remaining
+            if states.get(instance_id) in retry_states
+        ]
+        if not remaining:
+            return
+        if attempt == LIFECYCLE_MAX_ATTEMPTS:
+            on_exhausted(remaining, attempt)
+            return
+        retry_delay = max(1.0, delay) * (2 ** (attempt - 1))
+        if on_retry is not None:
+            on_retry(remaining, retry_delay)
+        time.sleep(retry_delay)
+
+
+def _resume_worker_batch(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    *,
+    stage: str,
+    detach: bool,
+    console: Console,
+) -> None:
+    """Resume one provider-sized batch, retrying only workers still stopped."""
+
+    def run_attempt(remaining: list[str]) -> bool:
+        try:
+            _run_lifecycle_command(
+                stage,
+                _resume_command(args, remaining, detach=detach),
+                console=console,
+                # Per-instance PMR output floods the materialization logs.
+                verbose=False,
+                live=False,
+            )
+        except PreparationError:
+            return False
+        return True
+
+    def on_exhausted(remaining: list[str], attempt: int) -> None:
+        raise PreparationError(
+            f"Could not resume {len(remaining):,} worker(s) in {stage} "
+            f"after {attempt:,} attempts: {', '.join(remaining)}"
+        )
+
+    def on_retry(remaining: list[str], retry_delay: float) -> None:
+        console.print(
+            Text.assemble(
+                ("↻", "bold yellow"),
+                f" {stage} incomplete; retrying {len(remaining):,} worker(s) ",
+                (f"in {retry_delay:g}s", "dim"),
+            )
+        )
+
+    _retry_lifecycle_batch(
+        args,
+        instance_ids,
+        run_attempt=run_attempt,
+        describe_states=lambda _remaining: _worker_states(args),
+        retry_states=frozenset({"stopped"}),
+        on_exhausted=on_exhausted,
+        on_retry=on_retry,
+    )
+
+
+def _resume_worker_groups_in_batches(
+    groups: Sequence[tuple[argparse.Namespace, Sequence[str]]],
+    *,
+    console: Console,
+) -> None:
+    """Resume every worker class in round-robin waves.
+
+    Wave size is the smallest batch size across the groups; the delay between
+    waves is the largest inter-batch delay across the groups.
+    """
+
+    if not groups:
+        return
+    batch_size = min(_provision_batch_size(group_args) for group_args, _ in groups)
+    delay = max(_provision_batch_delay(group_args) for group_args, _ in groups)
+    queues = deque(
+        (group_args, deque(sorted(set(instance_ids))))
+        for group_args, instance_ids in groups
+        if instance_ids
+    )
+    waves: list[list[tuple[argparse.Namespace, list[str]]]] = []
+    while queues:
+        wave_by_group: dict[int, tuple[argparse.Namespace, list[str]]] = {}
+        for _ in range(batch_size):
+            if not queues:
+                break
+            group_args, instance_queue = queues.popleft()
+            key = id(group_args)
+            if key not in wave_by_group:
+                wave_by_group[key] = (group_args, [])
+            wave_by_group[key][1].append(instance_queue.popleft())
+            if instance_queue:
+                queues.append((group_args, instance_queue))
+        waves.append(list(wave_by_group.values()))
+
+    for wave_index, wave in enumerate(waves, start=1):
+        with ThreadPoolExecutor(max_workers=len(wave)) as pool:
+            futures = [
+                pool.submit(
+                    _resume_worker_batch,
+                    group_args,
+                    instance_ids,
+                    stage=(
+                        f"resume batch {wave_index:,}/{len(waves):,} · "
+                        f"{_pmr_discovery_name(group_args)}"
+                    ),
+                    detach=True,
+                    console=console,
+                )
+                for group_args, instance_ids in wave
+            ]
+            for future in as_completed(futures):
+                future.result()
+        if wave_index < len(waves) and delay:
+            time.sleep(delay)
+
+
+def _pause_command(args: argparse.Namespace, instance_ids: Sequence[str]) -> list[str]:
+    options = _instance_options(args, instance_ids)
+    options.pop("ssh_key_path")
+    return build_poormanray_instance_command("pause", **options)
+
+
+def _worker_minimum_available_bytes(rows: Sequence[dict[str, str]]) -> int:
+    largest_working_set = max(int(row["estimated_peak_local_bytes"]) for row in rows)
+    # Integer ceiling of 110% of the largest working set: 10% headroom over the
+    # plan's estimate.
+    return (largest_working_set * 11 + 9) // 10
+
+
+def _storage_transfer_command(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> list[str]:
+    return build_poormanray_transfer_command(
+        transfers=((WORKER_STORAGE_SCRIPT, REMOTE_STORAGE_SCRIPT),),
+        **_instance_options(args, instance_ids),
+    )
+
+
+def _storage_setup_command(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    rows: Sequence[dict[str, str]],
+) -> list[str]:
+    minimum_bytes = _worker_minimum_available_bytes(rows)
+    remote_command = (
+        f"DOLMA_MIN_AVAILABLE_BYTES={minimum_bytes} "
+        f"bash {shlex.quote(REMOTE_STORAGE_SCRIPT)} --apply --layout {args.storage_layout}"
+    )
+    return build_poormanray_run_command(
+        remote_command=remote_command,
+        **_instance_options(args, instance_ids),
+    )
+
+
+def _runtime_setup_command(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> list[str]:
+    return build_poormanray_setup_dolma_command(**_instance_options(args, instance_ids))
+
+
+def _runtime_transfer_command(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> list[str]:
+    return build_poormanray_transfer_command(
+        transfers=(
+            (RESHARD_MODULE, REMOTE_RESHARD_MODULE),
+            (DOCUMENT_SELECTION_MODULE, REMOTE_DOCUMENT_SELECTION_MODULE),
+        ),
+        **_instance_options(args, instance_ids),
+    )
+
+
+def _runtime_validation_command(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> list[str]:
+    remote_command = (
+        "set -euo pipefail; export PYTHONSAFEPATH=1; cd /tmp; "
+        f'python_bin="{WORKER_PYTHON_BIN}"; '
+        'module_dir=$("$python_bin" -P -c \'import pathlib, dolma.tokenizer; '
+        "print(pathlib.Path(dolma.tokenizer.__file__).parent)'); "
+        'install -m 0644 /tmp/dolma3p5-runtime/reshard.py "$module_dir/reshard.py"; '
+        "install -m 0644 /tmp/dolma3p5-runtime/document_selection.py "
+        '"$module_dir/document_selection.py"; '
+        '"$python_bin" -P -c \'from dolma.tokenizer.reshard import '
+        "RESHARDING_MANIFEST_SCHEMA_VERSION; assert RESHARDING_MANIFEST_SCHEMA_VERSION == 2'; "
+        f"s5cmd version; findmnt {WORKER_LOCAL_MOUNT}; "
+        f"test -w {WORKER_LOCAL_TEMP_ROOT}"
+    )
+    return build_poormanray_run_command(
+        remote_command=remote_command,
+        **_instance_options(args, instance_ids),
+    )
+
+
+def _map_command(
+    args: argparse.Namespace,
+    script_dir: Path,
+    instance_ids: Sequence[str] = (),
+    *,
+    spindown: bool = True,
+) -> list[str]:
+    return build_poormanray_map_command(
+        cluster=_pmr_discovery_name(args),
+        project=args.project,
+        region=args.region,
+        script_dir=script_dir,
+        spindown=spindown,
+        instance_ids=instance_ids,
+        ssh_key_path=args.ssh_key_path,
+    )
+
+
+def _read_csv(path: Path) -> list[dict[str, str]]:
+    if path.is_symlink() or not path.is_file():
+        raise PreparationError(
+            f"Required execution artifact is missing or unsafe: {path}"
+        )
+    with path.open(newline="", encoding="utf-8") as handle:
+        return list(csv.DictReader(handle))
+
+
+def _load_execution_units(build: Path) -> list[dict[str, str]]:
+    _validate_preparation_build(build)
+    _validate_execution_layout(build)
+    rows = _read_csv(build / "01-plan/execution/config-index.csv")
+    if not rows:
+        raise PreparationError("Execution plan contains no units")
+    required = {
+        "unit_id",
+        "leaf_id",
+        "mix_name",
+        "category_name",
+        "launcher_path",
+        "destination_prefix",
+        "planned_uint32_values",
+        "planned_output_shard_count",
+        "estimated_peak_local_bytes",
+        "worker_instance_type",
+        "worker_storage_layout",
+        "worker_vcpus",
+        # Read only during post-run verification; checked here so a stale plan
+        # fails before workers are provisioned.
+        "allowed_materialized_target_residual_uint32_values",
+    }
+    missing = required - set(rows[0])
+    if missing:
+        raise PreparationError("Execution index is missing columns: " + ", ".join(sorted(missing)))
+    return rows
+
+
+def _planned_worker_groups(args: argparse.Namespace, rows: Sequence[dict[str, str]]) -> list[PlannedWorkerGroup]:
+    """Group selected units by planned worker instance type and storage layout."""
+
+    if args.instance_type:
+        if args.storage_layout == "auto":
+            planned_layouts = {
+                row["worker_storage_layout"] for row in rows if row["worker_instance_type"] == args.instance_type
+            }
+            planned_types = {row["worker_instance_type"] for row in rows}
+            if planned_types != {args.instance_type} or len(planned_layouts) != 1:
+                raise PreparationError(
+                    "--storage-layout must be explicit when --instance-type overrides " "the execution plan"
+                )
+            storage_layout = planned_layouts.pop()
+        else:
+            storage_layout = args.storage_layout
+        return [
+            PlannedWorkerGroup(
+                instance_type=args.instance_type,
+                storage_layout=storage_layout,
+                rows=tuple(rows),
+            )
+        ]
+
+    if args.storage_layout != "auto":
+        raise PreparationError("--storage-layout can only override the plan together with --instance-type")
+    grouped: dict[tuple[str, str, int], list[dict[str, str]]] = defaultdict(list)
+    for row in rows:
+        layout = row["worker_storage_layout"]
+        if layout not in {"single", "raid0"}:
+            raise PreparationError(f"Invalid planned storage layout for {row['unit_id']}: {layout}")
+        key = (
+            row["worker_instance_type"],
+            layout,
+            int(row["worker_vcpus"]),
+        )
+        grouped[key].append(row)
+    return [
+        PlannedWorkerGroup(instance_type, layout, tuple(group_rows))
+        for (instance_type, layout, _), group_rows in sorted(
+            grouped.items(), key=lambda item: (item[0][2], item[0][0])
+        )
+    ]
+
+
+def _worker_counts_for_groups(
+    groups: Sequence[PlannedWorkerGroup], maximum_workers: int
+) -> list[int]:
+    """Split maximum_workers across groups, giving every group at least one worker.
+
+    Each group is capped at its unit count and at PMR_DISCOVERY_WORKER_LIMIT.
+    Raises PreparationError when maximum_workers is below the group count.
+    """
+
+    if not groups:
+        return []
+    if maximum_workers < len(groups):
+        raise PreparationError(
+            f"--parallelism must be at least {len(groups):,} to run all planned "
+            "worker groups concurrently"
+        )
+
+    capacities = [
+        min(len(group.rows), PMR_DISCOVERY_WORKER_LIMIT) for group in groups
+    ]
+    total_capacity = sum(capacities)
+    remaining = min(maximum_workers, total_capacity) - len(groups)
+    counts = [1] * len(groups)
+    while remaining:
+        allocated = False
+        for index, group in enumerate(groups):
+            if counts[index] >= capacities[index]:
+                continue
+            counts[index] += 1
+            remaining -= 1
+            allocated = True
+            if not remaining:
+                break
+        if not allocated:
+            break
+    return counts
+
+
+def _category_selector(row: dict[str, str]) -> str:
+    return f"{row['mix_name']}::{row['category_name']}"
+
+
+def _category_rows(rows: Sequence[dict[str, str]]) -> list[dict[str, Any]]:
+    grouped: dict[tuple[str, str, str], list[dict[str, str]]] = defaultdict(list)
+    for row in rows:
+        grouped[(row["leaf_id"], row["mix_name"], row["category_name"])].append(row)
+    output: list[dict[str, Any]] = []
+    for (leaf_id, mix_name, category_name), units in grouped.items():
+        output.append(
+            {
+                "leaf_id": leaf_id,
+                "mix_name": mix_name,
+                "category_name": category_name,
+                "selector": f"{mix_name}::{category_name}",
+                "execution_units": len(units),
+                "planned_uint32_values": sum(int(row["planned_uint32_values"]) for row in units),
+                "planned_output_shard_count": sum(int(row["planned_output_shard_count"]) for row in units),
+                "largest_estimated_peak_local_bytes": max(int(row["estimated_peak_local_bytes"]) for row in units),
+                "worker_instance_types": ",".join(sorted({row["worker_instance_type"] for row in units})),
+            }
+        )
+    return sorted(output, key=lambda row: (int(row["leaf_id"].split(":", 1)[0]), row["leaf_id"]))
+
+
+def _print_categories(rows: Sequence[dict[str, str]], filter_text: str) -> None:
+    needle = filter_text.casefold()
+    matches = [
+        row
+        for row in _category_rows(rows)
+        if not needle
+        or needle in " ".join((row["leaf_id"], row["mix_name"], row["category_name"], row["selector"])).casefold()
+    ]
+    if not matches:
+        raise PreparationError(f"No categories match: {filter_text}")
+    print("leaf_id\texecution_units\toutput_shards\tplanned_tokens\tlargest_unit\tworkers\tselector")
+    for row in matches:
+        print(
+            f"{row['leaf_id']}\t{row['execution_units']:,}\t"
+            f"{row['planned_output_shard_count']:,}\t"
+            f"{_human_token_count(int(row['planned_uint32_values']))}\t"
+            f"{_human_byte_count(int(row['largest_estimated_peak_local_bytes']))}\t"
+            f"{row['worker_instance_types']}\t"
+            f"{row['selector']}"
+        )
+
+
+def _select_units(
+    args: argparse.Namespace, rows: Sequence[dict[str, str]]
+) -> tuple[str, list[dict[str, str]]]:
+    exclusions = list(getattr(args, "exclude_category", ()))
+    if exclusions and not args.all:
+        raise PreparationError("--exclude-category can only be used with --all")
+    if args.all:
+        excluded_unit_ids: set[str] = set()
+        for selector in exclusions:
+            excluded_unit_ids.update(
+                row["unit_id"]
+                for row in _filter_execution_units(rows, category=selector)
+            )
+        selected = [row for row in rows if row["unit_id"] not in excluded_unit_ids]
+        if not selected:
+            raise PreparationError("Category exclusions removed every execution unit")
+        label = "all" if not exclusions else f"all-except-{len(exclusions)}-categories"
+        return label, selected
+    if args.unit:
+        selected = _filter_execution_units(rows, unit=args.unit)
+        return f"unit-{args.unit}", selected
+
+    selector = str(args.category)
+    selected = _filter_execution_units(rows, category=selector)
+    return f"category-{selector}", selected
+
+
+def _safe_slug(value: str) -> str:
+    slug = "".join(
+        character.lower() if character.isalnum() else "-" for character in value
+    )
+    return "-".join(part for part in slug.split("-") if part)[:80] or "selection"
+
+
+def _resolve_launcher(build: Path, row: dict[str, str]) -> Path:
+    launcher = (build / row["launcher_path"]).resolve()
+    launcher_root = (build / "01-plan/execution/launcher-scripts").resolve()
+    if (
+        launcher.parent != launcher_root
+        or launcher.is_symlink()
+        or not launcher.is_file()
+    ):
+        raise PreparationError(
+            f"Unsafe or missing launcher for {row['unit_id']}: {launcher}"
+        )
+    return launcher
+
+
+def _safe_path_launcher_payload(
+    launcher: Path,
+    status_run_id: str | None = None,
+) -> bytes:
+    """Return the launcher script with the safe-path directives inserted.
+
+    Adds PYTHONSAFEPATH, `cd /tmp`, and DOLMA_STATUS_ROOT when status_run_id is
+    given. Raises PreparationError when the launcher has no `set -euo pipefail`
+    line or when status_run_id is not [a-z0-9-]+.
+    """
+
+    text = launcher.read_text(encoding="utf-8")
+    safe_path_export = "export PYTHONSAFEPATH=1"
+    safe_working_directory = "cd /tmp"
+    strict_mode = "set -euo pipefail\n"
+    if strict_mode not in text:
+        raise PreparationError(
+            f"Worker launcher is missing strict shell mode: {launcher}"
+        )
+    missing_directives = [
+        directive
+        for directive in (safe_path_export, safe_working_directory)
+        if directive not in text
+    ]
+    if missing_directives:
+        inserted = "\n".join(missing_directives)
+        text = text.replace(strict_mode, f"{strict_mode}\n{inserted}\n", 1)
+    if status_run_id is not None:
+        if not re.fullmatch(r"[a-z0-9-]+", status_run_id):
+            raise PreparationError(f"Unsafe materialization run ID: {status_run_id!r}")
+        status_directive = (
+            f'export DOLMA_STATUS_ROOT="{_remote_status_root(status_run_id)}"'
+        )
+        text = text.replace(strict_mode, f"{strict_mode}\n{status_directive}\n", 1)
+    return text.encode("utf-8")
+
+
+def _stage_selection(
+    build: Path,
+    label: str,
+    rows: Sequence[dict[str, str]],
+    status_run_id: str | None = None,
+) -> Path:
+    launchers = []
+    for row in rows:
+        launcher = _resolve_launcher(build, row)
+        payload = _safe_path_launcher_payload(launcher, status_run_id)
+        launchers.append((row, launcher, payload, hashlib.sha256(payload).digest()))
+    digest = hashlib.sha256()
+    for row, _, _, launcher_digest in sorted(
+        launchers, key=lambda item: item[0]["unit_id"]
+    ):
+        digest.update(row["unit_id"].encode())
+        digest.update(launcher_digest)
+    selection_name = f"{_safe_slug(label)}-{digest.hexdigest()[:12]}"
+    dispatch_root = build / "01-plan/execution/dispatch"
+    if dispatch_root.exists() and (
+        dispatch_root.is_symlink() or not dispatch_root.is_dir()
+    ):
+        raise PreparationError(f"Unsafe dispatch artifact path: {dispatch_root}")
+    dispatch_root.mkdir(exist_ok=True)
+    selection_dir = dispatch_root / selection_name
+    expected_names = {launcher.name for _, launcher, _, _ in launchers}
+    if selection_dir.exists():
+        if selection_dir.is_symlink() or not selection_dir.is_dir():
+            raise PreparationError(f"Unsafe dispatch selection path: {selection_dir}")
+        children = list(selection_dir.iterdir())
+        actual_names = {path.name for path in children}
+        unsafe = any(path.is_symlink() or not path.is_file() for path in children)
+        content_changed = any(
+            hashlib.sha256((selection_dir / launcher.name).read_bytes()).digest()
+            != launcher_digest
+            for _, launcher, _, launcher_digest in launchers
+            if (selection_dir / launcher.name).is_file()
+        )
+        if actual_names != expected_names or unsafe or content_changed:
+            raise PreparationError(
+                f"Existing dispatch selection does not match the plan: {selection_dir}"
+            )
+        return selection_dir
+
+    selection_dir.mkdir(exist_ok=False)
+    for _, launcher, payload, _ in launchers:
+        destination = selection_dir / launcher.name
+        destination.write_bytes(payload)
+        destination.chmod(launcher.stat().st_mode & 0o777)
+    return selection_dir
+
+
+def _stage_worker_assignments(
+    group: MaterializationGroup,
+) -> tuple[WorkerAssignment, ...]:
+    """Return one single-launcher poormanray directory per unit, largest unit first."""
+
+    assignments: list[WorkerAssignment] = []
+    assignment_root = group.script_dir.parent / f"{group.script_dir.name}-assignments"
+    if assignment_root.exists():
+        if assignment_root.is_symlink() or not assignment_root.is_dir():
+            raise PreparationError(
+                f"Unsafe worker-assignment path: {assignment_root}"
+            )
+    else:
+        assignment_root.mkdir(exist_ok=False)
+
+    ordered = sorted(
+        group.rows,
+        key=lambda row: (
+            int(row.get("estimated_work_uint32_values") or row["planned_uint32_values"]),
+            row["unit_id"],
+        ),
+        reverse=True,
+    )
+    for row in ordered:
+        unit_id = row["unit_id"]
+        if not re.fullmatch(r"[A-Za-z0-9._-]+", unit_id) or unit_id in {".", ".."}:
+            raise PreparationError(f"Unsafe execution-unit ID: {unit_id!r}")
+        launcher = group.script_dir / Path(row["launcher_path"]).name
+        if launcher.is_symlink() or not launcher.is_file():
+            raise PreparationError(
+                f"Staged execution-unit launcher is missing or unsafe: {launcher}"
+            )
+        assignment_dir = assignment_root / unit_id
+        destination = assignment_dir / launcher.name
+        if assignment_dir.exists():
+            if assignment_dir.is_symlink() or not assignment_dir.is_dir():
+                raise PreparationError(
+                    f"Unsafe worker-assignment directory: {assignment_dir}"
+                )
+            children = list(assignment_dir.iterdir())
+            if (
+                len(children) != 1
+                or children[0] != destination
+                or destination.is_symlink()
+                or not destination.is_file()
+                or destination.read_bytes() != launcher.read_bytes()
+                or (destination.stat().st_mode & 0o777)
+                != (launcher.stat().st_mode & 0o777)
+            ):
+                raise PreparationError(
+                    f"Existing worker assignment does not match the plan: {assignment_dir}"
+                )
+        else:
+            assignment_dir.mkdir(exist_ok=False)
+            destination.write_bytes(launcher.read_bytes())
+            destination.chmod(launcher.stat().st_mode & 0o777)
+        assignments.append(WorkerAssignment(group, (row,), assignment_dir))
+    return tuple(assignments)
+
+
+def _require_preflight(build: Path, selected: Sequence[dict[str, str]]) -> str:
+    summary_path = build / "02-preflight/preflight-summary.json"
+    if summary_path.is_symlink() or not summary_path.is_file():
+        raise PreparationError(
+            "A passing preflight is required before --execute. Add --preflight to this command "
+            "or run python scripts/dolma3p5_resharding/preflight.py immediately before dispatch."
+        )
+    try:
+        summary = json.loads(summary_path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise PreparationError(f"Invalid preflight summary: {summary_path}") from exc
+    blocking_fields = ("drifted_input_objects", "occupied_destinations", "errors")
+    if not summary.get("passed") or any(
+        int(summary.get(name, 0)) for name in blocking_fields
+    ):
+        raise PreparationError(f"Preflight did not pass: {summary_path}")
+    selected_digest = _unit_selection_digest(selected)
+    preflight_scope = summary.get("selection_scope")
+    if (
+        preflight_scope != "all"
+        and summary.get("selected_unit_ids_sha256") != selected_digest
+    ):
+        raise PreparationError(
+            "Preflight selection does not match this dispatch. Rerun preflight for the same "
+            "materialization selection."
+        )
+
+    destination_rows = _read_csv(build / "02-preflight/destination-status.csv")
+    statuses = {row["unit_id"]: row["status"] for row in destination_rows}
+    invalid = [
+        row["unit_id"] for row in selected if statuses.get(row["unit_id"]) != "empty"
+    ]
+    if invalid:
+        raise PreparationError(
+            f"Preflight does not show an empty destination for {len(invalid):,} selected unit(s)"
+        )
+    return str(summary.get("created_at", "unknown"))
+
+
+def _run_selected_preflight(
+    args: argparse.Namespace,
+    build: Path,
+    selected: Sequence[dict[str, str]],
+) -> None:
+    preflight_build(
+        argparse.Namespace(
+            build=build,
+            profile=args.profile,
+            region=args.region,
+            max_workers=None,
+            category=args.category,
+            unit=args.unit,
+            selected_unit_ids=tuple(row["unit_id"] for row in selected),
+            quiet=True,
+        )
+    )
+
+
+class _WorkerCleanupAbandoned(Exception):
+    """Raised after worker cleanup failed and the manual command was printed."""
+
+
+def _abandon_worker_cleanup(reason: str, command: Sequence[str]) -> NoReturn:
+    """Print the manual cleanup command and raise _WorkerCleanupAbandoned."""
+
+    print(
+        f"WARNING: {reason}; run {shlex.join(command)} immediately",
+        file=sys.stderr,
+    )
+    raise _WorkerCleanupAbandoned(reason)
+
+
+def _pause_worker_batch(
+    args: argparse.Namespace, instance_ids: Sequence[str], *, stage: str
+) -> None:
+    """Pause one provider-sized batch, retrying only workers still active."""
+
+    def run_attempt(remaining: list[str]) -> bool:
+        command = _pause_command(args, remaining)
+        try:
+            return_code = _run_compact_process(stage, command, verbose=False)
+        except OSError as exc:
+            _abandon_worker_cleanup(f"worker cleanup could not start: {exc}", command)
+        return return_code == 0
+
+    def describe_states(remaining: list[str]) -> dict[str, str]:
+        try:
+            return _worker_states(args)
+        except Exception as exc:
+            _abandon_worker_cleanup(
+                f"worker cleanup state check failed: {exc}",
+                _pause_command(args, remaining),
+            )
+
+    def on_exhausted(remaining: list[str], attempt: int) -> None:
+        _abandon_worker_cleanup(
+            f"worker cleanup failed after {attempt} attempts",
+            _pause_command(args, remaining),
+        )
+
+    _retry_lifecycle_batch(
+        args,
+        instance_ids,
+        run_attempt=run_attempt,
+        describe_states=describe_states,
+        retry_states=frozenset({"pending", "running"}),
+        on_exhausted=on_exhausted,
+    )
+
+
+def _pause_workers_after_failure(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> None:
+    if not instance_ids:
+        return
+    ordered = sorted(set(instance_ids))
+    delay = _provision_batch_delay(args)
+    batches = _lifecycle_batches(ordered, _provision_batch_size(args))
+    try:
+        for batch_index, batch in enumerate(batches, start=1):
+            _pause_worker_batch(
+                args,
+                batch,
+                stage=(
+                    f"pause workers after failure batch {batch_index}/{len(batches)}"
+                ),
+            )
+            if batch_index < len(batches) and delay:
+                time.sleep(delay)
+    except _WorkerCleanupAbandoned:
+        return
+
+
+def _prepare_workers(
+    args: argparse.Namespace,
+    worker_count: int,
+    *,
+    owned_instance_ids: Sequence[str] = (),
+    wait_for_ready: bool = True,
+    delay_after_last_batch: bool = False,
+    deferred_resume_ids: list[str] | None = None,
+) -> list[str]:
+    """Resume compatible stopped workers and create the missing ones in batches.
+
+    Raises PreparationError when the cluster holds active or transitioning
+    workers that are not in owned_instance_ids.
+    """
+
+    try:
+        before = _describe_cluster_instances(args.cluster, args.region, args.profile)
+    except Exception as exc:
+        raise PreparationError(
+            f"Could not inspect poormanray cluster {args.cluster!r} in {args.region}: {exc}"
+        ) from exc
+
+    owned = set(owned_instance_ids)
+    busy = [
+        instance
+        for instance in before
+        if instance.state != "stopped" and instance.instance_id not in owned
+    ]
+    if busy:
+        states = ", ".join(
+            f"{instance.instance_id}={instance.state}" for instance in busy
+        )
+        raise PreparationError(
+            f"Cluster {args.cluster!r} already has active or transitioning workers ({states}); "
+            "refusing to mix this materialization with another lifecycle"
+        )
+
+    reusable = [
+        instance
+        for instance in before
+        if instance.state == "stopped"
+        and instance.instance_type == args.instance_type
+        and instance.project == args.project
+    ][:worker_count]
+    selected_ids = [instance.instance_id for instance in reusable]
+    before_ids = {instance.instance_id for instance in before}
+    name_pattern = re.compile(rf"^{re.escape(args.cluster)}-(\d+)$")
+    existing_name_indices = [
+        int(match.group(1))
+        for instance in before
+        if (match := name_pattern.fullmatch(instance.name)) is not None
+    ]
+    next_name_index = max(existing_name_indices, default=-1) + 1
+
+    try:
+        if selected_ids:
+            _retag_cluster_instances(args, selected_ids)
+            if deferred_resume_ids is None:
+                _resume_workers_in_batches(args, selected_ids, detach=False)
+            else:
+                deferred_resume_ids.extend(selected_ids)
+
+        missing = worker_count - len(selected_ids)
+        batches = _provision_batches(missing, _provision_batch_size(args))
+        created_count = 0
+        for batch_index, batch_count in enumerate(batches, start=1):
+            _run_lifecycle_command(
+                f"create worker batch {batch_index}/{len(batches)}",
+                _create_command(args, batch_count, detach=True),
+                verbose=getattr(args, "verbose", False),
+            )
+            after = _describe_cluster_instances(args.cluster, args.region, args.profile)
+            created = [
+                instance
+                for instance in after
+                if instance.instance_id not in before_ids
+                and instance.state in {"pending", "running"}
+                and instance.instance_type == args.instance_type
+                and instance.project == args.project
+            ]
+            new_created = sorted(
+                instance.instance_id
+                for instance in created
+                if instance.instance_id not in selected_ids
+            )
+            if len(new_created) != batch_count:
+                selected_ids.extend(new_created)
+                raise PreparationError(
+                    f"Expected poormanray to create {batch_count:,} worker(s) in batch "
+                    f"{batch_index:,}/{len(batches):,}, but found {len(new_created):,}"
+                )
+            created_names = {
+                instance_id: (
+                    f"{args.cluster}-{next_name_index + created_count + offset:04d}"
+                )
+                for offset, instance_id in enumerate(new_created)
+            }
+            _retag_cluster_instances(args, new_created, names=created_names)
+            selected_ids.extend(new_created)
+            created_count += len(new_created)
+
+            should_delay = batch_index < len(batches) or (
+                delay_after_last_batch and batch_index == len(batches)
+            )
+            if should_delay:
+                time.sleep(_provision_batch_delay(args))
+
+        if len(selected_ids) != worker_count:
+            raise PreparationError(
+                f"Worker lifecycle selected {len(selected_ids):,} workers; expected {worker_count:,}"
+            )
+        if wait_for_ready:
+            _run_lifecycle_command(
+                "wait for workers",
+                _wait_command(args, selected_ids),
+                verbose=getattr(args, "verbose", False),
+            )
+        return sorted(selected_ids)
+    except BaseException:
+        if len(selected_ids) < worker_count:
+            try:
+                after_failure = _describe_cluster_instances(
+                    args.cluster, args.region, args.profile
+                )
+                selected_ids.extend(
+                    instance.instance_id
+                    for instance in after_failure
+                    if instance.instance_id not in before_ids
+                    and instance.state in {"pending", "running"}
+                    and instance.project == args.project
+                )
+            except (BotoCoreError, ClientError) as cleanup_exc:
+                print(
+                    f"WARNING: could not discover partially created workers for cleanup: {cleanup_exc}",
+                    file=sys.stderr,
+                )
+        try:
+            _retag_cluster_instances(args, sorted(set(selected_ids)))
+        except (BotoCoreError, ClientError, PreparationError) as tag_exc:
+            print(
+                f"WARNING: could not apply worker tags before cleanup: {tag_exc}",
+                file=sys.stderr,
+            )
+        _pause_workers_after_failure(args, sorted(set(selected_ids)))
+        raise
+
+
+def _worker_log_command(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    status_run_id: str,
+    log_offsets: dict[tuple[str, str], int] | None = None,
+    *,
+    include_logs: bool = True,
+) -> list[str]:
+    status_root = _remote_status_root(status_run_id)
+    active_instance_ids = set(instance_ids)
+    offset_payload = json.dumps(
+        {
+            log_name: offset
+            for (instance_id, log_name), offset in (log_offsets or {}).items()
+            if instance_id in active_instance_ids
+        },
+        sort_keys=True,
+        separators=(",", ":"),
+    )
+    log_reader = r'''import json
+import sys
+from pathlib import Path
+
+root = Path(sys.argv[1])
+offsets = json.loads(sys.argv[2])
+budget = int(sys.argv[3])
+for path in sorted(root.glob("*.log")):
+    if budget <= 0:
+        break
+    size = path.stat().st_size
+    offset = int(offsets.get(path.name, 0))
+    if offset < 0 or offset > size:
+        offset = 0
+    if offset == size:
+        continue
+    with path.open("rb") as handle:
+        handle.seek(offset)
+        data = handle.read(min(size - offset, budget))
+    if offset + len(data) < size:
+        boundary = max(data.rfind(b"\n"), data.rfind(b"\r"))
+        if boundary >= 0:
+            data = data[: boundary + 1]
+    if not data:
+        continue
+    end = offset + len(data)
+    print(f"@@DOLMA_LOG_BEGIN@@\t{path.name}\t{offset}\t{end}\t{size}")
+    text = data.decode("utf-8", errors="replace").replace("\r\n", "\n").replace("\r", "\n")
+    sys.stdout.write(text)
+    if not text.endswith("\n"):
+        sys.stdout.write("\n")
+    print(f"@@DOLMA_LOG_END@@\t{path.name}\t{end}\tOK")
+    budget -= len(data)
+'''
+    log_section = (
+        f'''python_bin="${{DOLMA_PYTHON:-{WORKER_PYTHON_BIN}}}"
+if [[ ! -x "$python_bin" ]]; then
+  python_bin=$(command -v python3.12 || command -v python3 || command -v python)
+fi
+"$python_bin" - "$status_root" {shlex.quote(offset_payload)} {WORKER_LOG_PAGE_BYTES} <<'PY'
+{log_reader}
+PY'''
+        if include_logs
+        else ""
+    )
+    remote_script = f"""status_root=\"{status_root}\"
+shopt -s nullglob
+status_files=(\"$status_root\"/*.status)
+if (( ${{#status_files[@]}} == 0 )); then
+  echo 'no unit status yet'
+else
+  for path in \"${{status_files[@]}}\"; do
+    printf '@@DOLMA_STATUS@@\t%s\t' \"$(basename \"$path\" .status)\"
+    tr -d '\n' < \"$path\"
+    printf '\n'
+  done
+fi
+{log_section}"""
+    return build_poormanray_run_command(
+        cluster=_pmr_discovery_name(args),
+        project=args.project,
+        region=args.region,
+        remote_command=f"bash -lc {shlex.quote(remote_script)}",
+        instance_ids=instance_ids,
+        parallelism=min(args.parallelism, len(instance_ids)),
+        ssh_key_path=args.ssh_key_path,
+    )
+
+
+def _worker_log_snapshots(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    status_run_id: str,
+    log_offsets: dict[tuple[str, str], int] | None = None,
+    *,
+    include_logs: bool = True,
+) -> dict[str, dict[str, Any]]:
+    """Return each worker's unit statuses and one page of new log bytes.
+
+    A page is at most WORKER_LOG_PAGE_BYTES. Returns an empty mapping when the
+    remote command fails or times out.
+    """
+
+    try:
+        result = subprocess.run(
+            _worker_log_command(
+                args,
+                instance_ids,
+                status_run_id,
+                log_offsets,
+                include_logs=include_logs,
+            ),
+            check=False,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            errors="replace",
+            timeout=WORKER_LOG_COMMAND_TIMEOUT_SECONDS,
+        )
+    except (OSError, subprocess.TimeoutExpired):
+        return {}
+    if result.returncode:
+        return {}
+
+    instance_payloads: dict[str, list[str]] = {}
+    current_instance: str | None = None
+    reading_stdout = False
+    for raw_line in result.stdout.splitlines():
+        line = _clean_process_line(raw_line)
+        instance_match = re.fullmatch(r"Instance ([^:]+):", line)
+        if instance_match:
+            current_instance = instance_match.group(1)
+            instance_payloads.setdefault(current_instance, [])
+            reading_stdout = False
+            continue
+        if line.startswith("stdout:"):
+            reading_stdout = True
+            remainder = line.removeprefix("stdout:").strip()
+            if remainder and current_instance is not None:
+                instance_payloads[current_instance].append(remainder)
+            continue
+        if line == "stderr:":
+            reading_stdout = False
+            continue
+        if reading_stdout and line and current_instance is not None:
+            instance_payloads[current_instance].append(line)
+
+    snapshots: dict[str, dict[str, Any]] = {}
+    for instance_id, payload in instance_payloads.items():
+        statuses: dict[str, str] = {}
+        logs: dict[str, tuple[str, ...]] = {}
+        offsets: dict[str, int] = {}
+        current_log: str | None = None
+        current_lines: list[str] = []
+        for line in payload:
+            if line.startswith("@@DOLMA_STATUS@@\t"):
+                _, unit_id, status = line.split("\t", 2)
+                statuses[unit_id] = status
+                continue
+            if line.startswith("@@DOLMA_LOG_BEGIN@@\t"):
+                current_log = line.split("\t", 2)[1]
+                current_lines = []
+                continue
+            if line.startswith("@@DOLMA_LOG_END@@\t"):
+                fields = line.split("\t")
+                if (
+                    current_log is not None
+                    and len(fields) == 4
+                    and fields[1] == current_log
+                    and fields[3] == "OK"
+                ):
+                    logs[current_log] = tuple(current_lines)
+                    offsets[current_log] = int(fields[2])
+                elif current_log is not None and len(fields) == 2:
+                    # Envelope without acknowledged byte offsets.
+                    logs[current_log] = tuple(current_lines)
+                current_log = None
+                current_lines = []
+                continue
+            if current_log is not None:
+                current_lines.append(line)
+        snapshots[instance_id] = {
+            "statuses": statuses,
+            "logs": logs,
+            "offsets": offsets,
+        }
+    return snapshots
+
+
+@dataclass(frozen=True)
+class _MonitorChannels:
+    """Shared state the warm-worker dispatcher and the progress monitor exchange.
+
+    Every field is optional; the monitor runs standalone when none are set.
+    """
+
+    worker_stages: dict[str, str] | None = None
+    stage_lock: Lock | None = None
+    all_dispatched: Event | None = None
+    abort: Event | None = None
+    worker_args: dict[str, argparse.Namespace] | None = None
+    unit_statuses: dict[tuple[str, str], str] | None = None
+    unit_status_lock: Lock | None = None
+    status_changed: Event | None = None
+
+    @property
+    def tracks_stages(self) -> bool:
+        return self.worker_stages is not None
+
+    @property
+    def tracks_units(self) -> bool:
+        return self.unit_statuses is not None
+
+    @property
+    def aborted(self) -> bool:
+        return self.abort is not None and self.abort.is_set()
+
+    @property
+    def dispatch_complete(self) -> bool:
+        return self.all_dispatched is None or self.all_dispatched.is_set()
+
+    def stage_guard(self) -> AbstractContextManager[Any]:
+        """Return the dispatcher's stage lock, or a null context when standalone."""
+
+        return self.stage_lock if self.stage_lock is not None else nullcontext()
+
+    def status_guard(self) -> AbstractContextManager[Any]:
+        """Return the dispatcher's unit-status lock, or a null context when standalone."""
+
+        return (
+            self.unit_status_lock
+            if self.unit_status_lock is not None
+            else nullcontext()
+        )
+
+
+@dataclass
+class _WorkerOutputTracker:
+    """Remember which worker statuses and log bytes have already been printed."""
+
+    output: Console
+    verbose: bool
+    worker_tags: dict[str, tuple[str, str]]
+    previous_statuses: dict[tuple[str, str], str] = field(default_factory=dict)
+    emitted_log_lines: dict[tuple[str, str], int] = field(default_factory=dict)
+    log_byte_offsets: dict[tuple[str, str], int] = field(default_factory=dict)
+    announced_logs: set[tuple[str, str]] = field(default_factory=set)
+
+    def record_statuses(
+        self,
+        channels: _MonitorChannels,
+        instance_id: str,
+        statuses: dict[str, str],
+    ) -> bool:
+        """Publish one worker's unit statuses and return whether any changed."""
+
+        worker_tag, worker_style = self.worker_tags[instance_id]
+        changed = False
+        for unit_id, status in sorted(statuses.items()):
+            assert isinstance(status, str)
+            status_key = (instance_id, unit_id)
+            if channels.unit_statuses is not None:
+                with channels.status_guard():
+                    if channels.unit_statuses.get(status_key) != status:
+                        channels.unit_statuses[status_key] = status
+                        changed = True
+            if self.verbose and self.previous_statuses.get(status_key) != status:
+                self.previous_statuses[status_key] = status
+                self.output.print(
+                    _worker_log_line(
+                        worker_tag,
+                        worker_style,
+                        f"{unit_id} · {status}",
+                        bold=True,
+                    )
+                )
+        return changed
+
+    def print_new_log_lines(self, instance_id: str, snapshot: dict[str, Any]) -> None:
+        """Print the log lines this monitor has not already shown."""
+
+        worker_tag, worker_style = self.worker_tags[instance_id]
+        logs = snapshot["logs"]
+        assert isinstance(logs, dict)
+        offsets = snapshot.get("offsets", {})
+        assert isinstance(offsets, dict)
+        for log_name, log_lines in sorted(logs.items()):
+            assert isinstance(log_lines, tuple)
+            log_key = (instance_id, log_name)
+            if log_name in offsets:
+                new_lines = log_lines
+            else:
+                # Envelope without acknowledged byte offsets: fall back to the
+                # emitted line count.
+                emitted = self.emitted_log_lines.get(log_key, 0)
+                if len(log_lines) < emitted:
+                    emitted = 0
+                new_lines = log_lines[emitted:]
+            if new_lines:
+                if log_key not in self.announced_logs:
+                    self.announced_logs.add(log_key)
+                    self.output.print(
+                        _worker_log_line(
+                            worker_tag,
+                            worker_style,
+                            f"unit {Path(log_name).stem}",
+                            bold=True,
+                        )
+                    )
+                for line in new_lines:
+                    timestamp, message = _worker_log_parts(line)
+                    self.output.print(
+                        _worker_log_line(
+                            worker_tag,
+                            worker_style,
+                            message,
+                            timestamp=timestamp,
+                        )
+                    )
+            if log_name in offsets:
+                self.log_byte_offsets[log_key] = int(offsets[log_name])
+            else:
+                self.emitted_log_lines[log_key] = len(log_lines)
+
+
+def _monitored_workers(
+    describe_instances: Callable[[str, str, str | None], list[ClusterInstance]],
+    args: argparse.Namespace,
+    expected_ids: set[str],
+) -> dict[str, ClusterInstance]:
+    """Return the monitored workers, raising PreparationError if any is missing."""
+
+    selected = {
+        instance.instance_id: instance
+        for instance in describe_instances(args.cluster, args.region, args.profile)
+        if instance.instance_id in expected_ids
+    }
+    missing = expected_ids - set(selected)
+    if missing:
+        raise PreparationError(
+            "Could not find selected materialization worker(s): "
+            + ", ".join(sorted(missing))
+        )
+    return selected
+
+
+def _reconcile_worker_stages(
+    channels: _MonitorChannels, selected: dict[str, ClusterInstance]
+) -> dict[str, str]:
+    """Fold provider stop states into the shared stage map, then snapshot it."""
+
+    stages = channels.worker_stages
+    assert stages is not None
+    with channels.stage_guard():
+        for instance_id, instance in selected.items():
+            base = stages.get(instance_id, "waiting").split(":", 1)[0]
+            if base in {"materializing", "stopping"} and instance.state in {
+                "stopping",
+                "stopped",
+            }:
+                stages[instance_id] = instance.state
+        return dict(stages)
+
+
+def _unit_progress_detail(channels: _MonitorChannels, unit_count: int) -> str:
+    """Return the unit-completion detail from the statuses the dispatcher published."""
+
+    if channels.unit_statuses is None:
+        return f"{unit_count:,} units"
+    with channels.status_guard():
+        completed_units = sum(
+            status == "succeeded" for status in channels.unit_statuses.values()
+        )
+    return f"{completed_units:,}/{unit_count:,} units"
+
+
+def _lifecycle_progress_detail(
+    channels: _MonitorChannels,
+    lifecycle_counts: Counter[str],
+    unit_count: int,
+    started_at: float,
+) -> str:
+    """Summarize the dispatcher's per-worker stages."""
+
+    return " · ".join(
+        (
+            _unit_progress_detail(channels, unit_count),
+            f"{lifecycle_counts['waiting']:,} waiting",
+            f"{lifecycle_counts['queued']:,} queued",
+            f"{lifecycle_counts['bootstrapping']:,} bootstrapping",
+            f"{lifecycle_counts['materializing']:,} materializing",
+            f"{lifecycle_counts['stopping']:,} stopping",
+            f"{lifecycle_counts['stopped']:,} stopped",
+            f"{lifecycle_counts['failed']:,} failed",
+            _elapsed_time(started_at),
+        )
+    )
+
+
+def _provider_progress_detail(
+    state_counts: Counter[str], unit_count: int, started_at: float
+) -> str:
+    """Summarize provider instance states when no dispatcher stages exist."""
+
+    return " · ".join(
+        (
+            f"{unit_count:,} units",
+            f"{state_counts['running']:,} running",
+            f"{state_counts['stopping']:,} stopping",
+            f"{state_counts['stopped']:,} stopped",
+            _elapsed_time(started_at),
+        )
+    )
+
+
+def _materializing_worker_ids(
+    channels: _MonitorChannels,
+    selected: dict[str, ClusterInstance],
+    stage_snapshot: dict[str, str],
+) -> list[str]:
+    """Return the running workers currently in the materializing stage."""
+
+    return sorted(
+        instance.instance_id
+        for instance in selected.values()
+        if instance.state == "running"
+        and (
+            not channels.tracks_stages
+            or stage_snapshot.get(instance.instance_id, "").split(":", 1)[0]
+            == "materializing"
+        )
+    )
+
+
+def _collect_worker_snapshots(
+    args: argparse.Namespace,
+    channels: _MonitorChannels,
+    instance_ids: Sequence[str],
+    status_run_id: str,
+    log_offsets: dict[tuple[str, str], int],
+    *,
+    include_logs: bool,
+) -> dict[str, dict[str, Any]]:
+    """Read one status and log page per poormanray discovery group."""
+
+    log_groups: dict[str, tuple[argparse.Namespace, list[str]]] = {}
+    for instance_id in instance_ids:
+        log_args = (channels.worker_args or {}).get(instance_id, args)
+        selector = _pmr_discovery_name(log_args)
+        if selector not in log_groups:
+            log_groups[selector] = (log_args, [])
+        log_groups[selector][1].append(instance_id)
+    snapshots: dict[str, dict[str, Any]] = {}
+    for log_args, group_instance_ids in log_groups.values():
+        snapshots.update(
+            _worker_log_snapshots(
+                log_args,
+                group_instance_ids,
+                status_run_id,
+                log_offsets,
+                include_logs=include_logs,
+            )
+        )
+    return snapshots
+
+
+def _publish_worker_output(
+    args: argparse.Namespace,
+    channels: _MonitorChannels,
+    tracker: _WorkerOutputTracker,
+    running_ids: Sequence[str],
+    status_run_id: str,
+) -> None:
+    """Print new worker output and publish new unit statuses to the dispatcher."""
+
+    snapshots = _collect_worker_snapshots(
+        args,
+        channels,
+        running_ids,
+        status_run_id,
+        tracker.log_byte_offsets,
+        include_logs=tracker.verbose,
+    )
+    statuses_changed = False
+    for instance_id in sorted(snapshots):
+        snapshot = snapshots[instance_id]
+        statuses = snapshot["statuses"]
+        assert isinstance(statuses, dict)
+        if tracker.record_statuses(channels, instance_id, statuses):
+            statuses_changed = True
+        if tracker.verbose:
+            tracker.print_new_log_lines(instance_id, snapshot)
+    if statuses_changed and channels.status_changed is not None:
+        channels.status_changed.set()
+
+
+def _await_next_worker_poll(
+    args: argparse.Namespace,
+    channels: _MonitorChannels,
+    sleep: Callable[[float], None],
+) -> None:
+    """Wait out one poll interval, waking early when the dispatcher aborts."""
+
+    poll_seconds = args.completion_poll_seconds
+    if channels.tracks_units:
+        poll_seconds = min(poll_seconds, args.readiness_poll_seconds)
+    if channels.abort is not None:
+        channels.abort.wait(poll_seconds)
+    else:
+        sleep(poll_seconds)
+
+
+def _wait_for_workers_to_stop(
+    args: argparse.Namespace,
+    instance_ids: Sequence[str],
+    unit_count: int,
+    status_run_id: str,
+    *,
+    describe: Callable[[str, str, str | None], list[ClusterInstance]] | None = None,
+    sleep: Callable[[float], None] = time.sleep,
+    console: Console | None = None,
+    worker_stages: dict[str, str] | None = None,
+    stage_lock: Lock | None = None,
+    all_dispatched: Event | None = None,
+    abort: Event | None = None,
+    worker_args: dict[str, argparse.Namespace] | None = None,
+    unit_statuses: dict[tuple[str, str], str] | None = None,
+    unit_status_lock: Lock | None = None,
+    status_changed: Event | None = None,
+) -> None:
+    """Monitor worker stages until every dispatched worker has stopped.
+
+    Each optional keyword is one shared-state channel to the dispatcher.
+    """
+
+    channels = _MonitorChannels(
+        worker_stages=worker_stages,
+        stage_lock=stage_lock,
+        all_dispatched=all_dispatched,
+        abort=abort,
+        worker_args=worker_args,
+        unit_statuses=unit_statuses,
+        unit_status_lock=unit_status_lock,
+        status_changed=status_changed,
+    )
+    describe_instances = describe or _describe_cluster_instances
+    output = console or Console(stderr=True, highlight=False)
+    expected_ids = set(instance_ids)
+    started_at = time.monotonic()
+    verbose = getattr(args, "verbose", False)
+    previous_state_signature: tuple[tuple[str, int], ...] | None = None
+    aborted = False
+    tracker = _WorkerOutputTracker(
+        output=output,
+        verbose=verbose,
+        worker_tags={
+            instance_id: (
+                _worker_tag(instance_id),
+                WORKER_LOG_STYLES[(index - 1) % len(WORKER_LOG_STYLES)],
+            )
+            for index, instance_id in enumerate(sorted(expected_ids), start=1)
+        },
+    )
+    live_status = (
+        output.status(_stage_status("workers"), spinner="dots")
+        if output.is_terminal
+        else None
+    )
+    if live_status is None:
+        output.print(Text.assemble(("…", "cyan"), " ", ("workers", "bold")))
+    else:
+        live_status.start()
+
+    try:
+        while True:
+            if channels.aborted:
+                aborted = True
+                break
+            selected = _monitored_workers(describe_instances, args, expected_ids)
+            state_counts = Counter(instance.state for instance in selected.values())
+            stopped = state_counts["stopped"]
+            lifecycle_counts: Counter[str] | None = None
+            stage_snapshot: dict[str, str] = {}
+            if channels.tracks_stages:
+                stage_snapshot = _reconcile_worker_stages(channels, selected)
+                lifecycle_counts = Counter(
+                    value.split(":", 1)[0] for value in stage_snapshot.values()
+                )
+                detail = _lifecycle_progress_detail(
+                    channels, lifecycle_counts, unit_count, started_at
+                )
+            else:
+                detail = _provider_progress_detail(state_counts, unit_count, started_at)
+            if live_status is not None:
+                live_status.update(_stage_status("workers", detail))
+            else:
+                state_signature = tuple(
+                    sorted((lifecycle_counts or state_counts).items())
+                )
+                if state_signature != previous_state_signature:
+                    output.print(Text(f"workers  {detail}", style="dim"))
+                    previous_state_signature = state_signature
+            running_ids = _materializing_worker_ids(channels, selected, stage_snapshot)
+            if running_ids and (verbose or channels.tracks_units):
+                _publish_worker_output(
+                    args, channels, tracker, running_ids, status_run_id
+                )
+            if channels.dispatch_complete and stopped == len(expected_ids):
+                break
+            _await_next_worker_poll(args, channels, sleep)
+    finally:
+        if live_status is not None:
+            live_status.stop()
+
+    if not aborted:
+        output.print(
+            Text.assemble(
+                ("✓", "bold green"),
+                " materialization workers stopped",
+                (f"  {_elapsed_time(started_at)}", "dim"),
+            )
+        )
+
+
+def _check_materialized_unit(client: Any, row: dict[str, str]) -> MaterializedUnitCheck:
+    parsed = urlparse(row["destination_prefix"])
+    if parsed.scheme != "s3" or not parsed.netloc:
+        return MaterializedUnitCheck(
+            row["unit_id"],
+            0,
+            0,
+            0,
+            (f"unsupported destination: {row['destination_prefix']}",),
+        )
+    prefix = parsed.path.lstrip("/").rstrip("/") + "/"
+    objects = _list_prefix(client, parsed.netloc, prefix)
+    object_map = {obj.key: obj for obj in objects}
+    npys = [obj for obj in objects if obj.key.endswith(".npy")]
+    metadata = {obj.key for obj in objects if obj.key.endswith(".csv.gz")}
+    problems: list[str] = []
+    if not npys:
+        problems.append("no NPY output")
+    planned_output_shards = int(row["planned_output_shard_count"])
+    if len(npys) != planned_output_shards:
+        problems.append(
+            f"found {len(npys):,} output shards; expected {planned_output_shards:,}"
+        )
+    invalid_npys = [
+        obj for obj in npys if obj.size_bytes <= 0 or obj.size_bytes % UINT32_BYTES
+    ]
+    if invalid_npys:
+        problems.append(f"{len(invalid_npys):,} invalid NPY sizes")
+    missing_metadata = [
+        _pair_metadata_key(obj.key)
+        for obj in npys
+        if _pair_metadata_key(obj.key) not in metadata
+    ]
+    if missing_metadata:
+        problems.append(f"{len(missing_metadata):,} NPYs missing metadata")
+    orphan_metadata = [
+        key for key in metadata if key[: -len(".csv.gz")] + ".npy" not in object_map
+    ]
+    if orphan_metadata:
+        problems.append(f"{len(orphan_metadata):,} orphan metadata files")
+    unexpected = [obj for obj in objects if not obj.key.endswith((".npy", ".csv.gz"))]
+    if unexpected:
+        problems.append(f"{len(unexpected):,} unexpected output objects")
+
+    actual = sum(obj.size_bytes // UINT32_BYTES for obj in npys)
+    predicted = int(row["planned_uint32_values"])
+    allowed = int(row["allowed_materialized_target_residual_uint32_values"])
+    residual = actual - predicted
+    if abs(residual) > allowed:
+        problems.append(
+            f"token estimate differs by {residual:+,}; allowed residual is {allowed:,}"
+        )
+    return MaterializedUnitCheck(
+        unit_id=row["unit_id"],
+        actual_uint32_values=actual,
+        npy_count=len(npys),
+        metadata_count=len(metadata),
+        problems=tuple(problems),
+    )
+
+
+def _verify_materialized_units(
+    args: argparse.Namespace,
+    rows: Sequence[dict[str, str]],
+    *,
+    client: Any | None = None,
+    console: Console | None = None,
+) -> list[MaterializedUnitCheck]:
+    """Check every selected unit's destination objects and their sizes.
+
+    Raises PreparationError when any unit has problems or its listing failed.
+    """
+
+    if client is None:
+        session = boto3.Session(profile_name=args.profile, region_name=args.region)
+        client = session.client("s3", region_name=args.region)
+    output = console or Console(stderr=True, highlight=False)
+    started_at = time.monotonic()
+    live_status = (
+        output.status(_stage_status("verify materialized outputs"), spinner="dots")
+        if output.is_terminal
+        else None
+    )
+    if live_status is None:
+        output.print(
+            Text.assemble(("…", "cyan"), " ", ("verify materialized outputs", "bold"))
+        )
+    else:
+        live_status.start()
+
+    checks: list[MaterializedUnitCheck] = []
+    request_errors: list[str] = []
+    try:
+        with ThreadPoolExecutor(max_workers=min(args.parallelism, len(rows))) as pool:
+            futures = {
+                pool.submit(_check_materialized_unit, client, row): row for row in rows
+            }
+            for completed, future in enumerate(as_completed(futures), start=1):
+                row = futures[future]
+                try:
+                    checks.append(future.result())
+                except Exception as exc:
+                    request_errors.append(f"{row['unit_id']}: {exc}")
+                if live_status is not None:
+                    live_status.update(
+                        _stage_status(
+                            "verify materialized outputs",
+                            f"{completed}/{len(rows)} units checked",
+                        )
+                    )
+    finally:
+        if live_status is not None:
+            live_status.stop()
+
+    failed = [check for check in checks if check.problems]
+    if request_errors or failed or len(checks) != len(rows):
+        details = [*request_errors]
+        details.extend(
+            f"{check.unit_id}: {', '.join(check.problems)}" for check in failed
+        )
+        summary = "; ".join(details[:5])
+        if len(details) > 5:
+            summary += f"; and {len(details) - 5:,} more"
+        raise PreparationError(
+            f"Materialization output verification failed for "
+            f"{len(request_errors) + len(failed):,} unit(s): {summary}"
+        )
+
+    total = sum(check.actual_uint32_values for check in checks)
+    output.print(
+        Text.assemble(
+            ("✓", "bold green"),
+            f" materialization verified  {len(checks):,}/{len(rows):,} units · ",
+            (_human_token_count(total), "bold"),
+            " tokens",
+            (f"  {_elapsed_time(started_at)}", "dim"),
+        )
+    )
+    return sorted(checks, key=lambda check: check.unit_id)
+
+
+def _dry_run_lifecycle_commands(
+    args: argparse.Namespace,
+    rows: Sequence[dict[str, str]],
+    script_dir: Path,
+    worker_count: int,
+) -> list[tuple[str, list[str]]]:
+    """Return the lifecycle commands for the create path.
+
+    --execute resumes compatible stopped workers when the cluster has any,
+    instead of creating them.
+    """
+
+    batches = _provision_batches(worker_count, _provision_batch_size(args))
+    create_commands = [
+        (
+            f"create worker batch {index}/{len(batches)}",
+            _create_command(args, count, detach=True),
+        )
+        for index, count in enumerate(batches, start=1)
+    ]
+    return [
+        *create_commands,
+        ("upload storage setup", _storage_transfer_command(args, ())),
+        ("prepare local NVMe", _storage_setup_command(args, (), rows)),
+        ("install Dolma and s5cmd", _runtime_setup_command(args, ())),
+        ("upload resharding runtime", _runtime_transfer_command(args, ())),
+        (
+            "install and validate resharding runtime",
+            _runtime_validation_command(args, ()),
+        ),
+        ("dispatch and stop workers when done", _map_command(args, script_dir)),
+    ]
+
+
+def _print_dispatch(
+    label: str,
+    rows: Sequence[dict[str, str]],
+    script_dir: Path,
+    lifecycle_commands: Sequence[tuple[str, Sequence[str]]],
+    worker_count: int,
+    execute: bool,
+    *,
+    cluster: str | None = None,
+    project: str | None = None,
+    region: str | None = None,
+    instance_type: str | None = None,
+    storage_layout: str | None = None,
+) -> None:
+    category_count = len({row["leaf_id"] for row in rows})
+    planned_tokens = sum(int(row["planned_uint32_values"]) for row in rows)
+    planned_output_shards = sum(int(row["planned_output_shard_count"]) for row in rows)
+    largest_unit = max(int(row["estimated_peak_local_bytes"]) for row in rows)
+    display_label = label.removeprefix("category-") if label.startswith("category-") else label
+
+    summary = Table.grid(padding=(0, 2))
+    summary.add_column(style="dim", no_wrap=True)
+    summary.add_column()
+    summary.add_row("Selection", display_label)
+    summary.add_row(
+        "Work",
+        " · ".join(
+            (
+                _count_label(category_count, "category", "categories"),
+                _count_label(len(rows), "unit"),
+                _count_label(worker_count, "worker"),
+            )
+        ),
+    )
+    summary.add_row(
+        "Output",
+        f"{_human_token_count(planned_tokens)} tokens · {planned_output_shards:,} shards",
+    )
+    summary.add_row("Largest working set", _human_byte_count(largest_unit))
+    if instance_type:
+        summary.add_row("Worker", instance_type)
+    if storage_layout:
+        summary.add_row("Local storage", storage_layout)
+    if cluster:
+        summary.add_row("Cluster", cluster)
+    if project:
+        summary.add_row("Project", project)
+    if region:
+        summary.add_row("Region", region)
+    output = Console(highlight=False)
+    output.print(Text("Execution" if execute else "Dry run", style="bold"))
+    output.print(summary)
+    if execute:
+        return
+    print("\ncommands:")
+    for _, command in lifecycle_commands:
+        print(shlex.join(command))
+    print("\nunits:")
+    for row in sorted(rows, key=lambda item: item["unit_id"]):
+        print(
+            f"{row['unit_id']}\t{_human_token_count(int(row['planned_uint32_values']))} tokens\t"
+            f"{int(row['planned_output_shard_count']):,} shards\t"
+            f"{_human_byte_count(int(row['estimated_peak_local_bytes']))}\t"
+            f"{row['destination_prefix']}"
+        )
+
+
+def _ready_worker_ids(
+    args: argparse.Namespace, instance_ids: Sequence[str]
+) -> set[str]:
+    """Return workers whose provider health checks have passed."""
+
+    if not instance_ids:
+        return set()
+    session = boto3.Session(profile_name=args.profile, region_name=args.region)
+    client = session.client("ec2", region_name=args.region)
+    ready: set[str] = set()
+    ordered = sorted(set(instance_ids))
+    for batch in _lifecycle_batches(ordered, DESCRIBE_INSTANCE_STATUS_ID_LIMIT):
+        response = client.describe_instance_status(
+            InstanceIds=batch,
+            IncludeAllInstances=True,
+        )
+        for status in response.get("InstanceStatuses", []):
+            if (
+                status.get("InstanceState", {}).get("Name") == "running"
+                and status.get("InstanceStatus", {}).get("Status") == "ok"
+                and status.get("SystemStatus", {}).get("Status") == "ok"
+            ):
+                instance_id = status.get("InstanceId")
+                if isinstance(instance_id, str):
+                    ready.add(instance_id)
+    return ready
+
+
+def _set_worker_stage(
+    stages: dict[str, str], lock: Lock, instance_id: str, stage: str
+) -> None:
+    with lock:
+        stages[instance_id] = stage
+
+
+def _bootstrap_and_dispatch_worker(
+    args: argparse.Namespace,
+    instance_id: str,
+    assignment: WorkerAssignment,
+    stages: dict[str, str],
+    stage_lock: Lock,
+    console: Console | None = None,
+) -> None:
+    """Prepare one ready worker and dispatch its assigned execution units."""
+
+    tag = _worker_tag(instance_id)
+
+    def run(stage: str, command: Sequence[str]) -> None:
+        _set_worker_stage(stages, stage_lock, instance_id, f"bootstrapping:{stage}")
+        _run_lifecycle_command(
+            f"{tag} · {stage}",
+            command,
+            console=console,
+            verbose=args.verbose,
+            live=False,
+        )
+
+    group_args = assignment.group.args
+    try:
+        run("upload storage setup", _storage_transfer_command(group_args, [instance_id]))
+        run(
+            "prepare local NVMe",
+            _storage_setup_command(group_args, [instance_id], assignment.group.rows),
+        )
+        run("install Dolma and s5cmd", _runtime_setup_command(group_args, [instance_id]))
+        run(
+            "upload resharding runtime",
+            _runtime_transfer_command(group_args, [instance_id]),
+        )
+        run(
+            "validate resharding runtime",
+            _runtime_validation_command(group_args, [instance_id]),
+        )
+        run(
+            f"dispatch {len(assignment.rows):,} unit(s)",
+            _map_command(
+                group_args,
+                assignment.script_dir,
+                [instance_id],
+                spindown=False,
+            ),
+        )
+        _set_worker_stage(stages, stage_lock, instance_id, "materializing")
+    except BaseException:
+        _set_worker_stage(stages, stage_lock, instance_id, "failed")
+        raise
+
+
+def _dispatch_assignment_to_worker(
+    instance_id: str,
+    assignment: WorkerAssignment,
+    stages: dict[str, str],
+    stage_lock: Lock,
+    console: Console,
+) -> None:
+    """Dispatch one unit to a warm worker without bootstrapping or stopping it."""
+
+    unit_id = assignment.rows[0]["unit_id"]
+    _set_worker_stage(stages, stage_lock, instance_id, f"queued:{unit_id}")
+    try:
+        _run_lifecycle_command(
+            f"{_worker_tag(instance_id)} · dispatch {unit_id}",
+            _map_command(
+                assignment.group.args,
+                assignment.script_dir,
+                [instance_id],
+                spindown=False,
+            ),
+            console=console,
+            verbose=assignment.group.args.verbose,
+            live=False,
+        )
+        _set_worker_stage(stages, stage_lock, instance_id, "materializing")
+    except BaseException:
+        _set_worker_stage(stages, stage_lock, instance_id, "failed")
+        raise
+
+
+def _stop_worker_after_work(
+    instance_id: str,
+    group_args: argparse.Namespace,
+    stages: dict[str, str],
+    stage_lock: Lock,
+    console: Console,
+) -> None:
+    """Stop one warm worker once its group's unit queue is empty."""
+
+    _set_worker_stage(stages, stage_lock, instance_id, "stopping")
+    try:
+        _run_lifecycle_command(
+            f"{_worker_tag(instance_id)} · stop",
+            _pause_command(group_args, [instance_id]),
+            console=console,
+            verbose=False,
+            live=False,
+        )
+    except BaseException:
+        _set_worker_stage(stages, stage_lock, instance_id, "failed")
+        raise
+
+
+@dataclass(frozen=True)
+class _DispatchChannels:
+    """Shared state the dispatcher owns and the progress monitor reads or fills."""
+
+    worker_stages: dict[str, str]
+    stage_lock: Lock = field(default_factory=Lock)
+    unit_statuses: dict[tuple[str, str], str] = field(default_factory=dict)
+    unit_status_lock: Lock = field(default_factory=Lock)
+    status_changed: Event = field(default_factory=Event)
+    all_dispatched: Event = field(default_factory=Event)
+    abort: Event = field(default_factory=Event)
+
+    def set_stage(self, instance_id: str, stage: str) -> None:
+        _set_worker_stage(self.worker_stages, self.stage_lock, instance_id, stage)
+
+    def stage_snapshot(self) -> dict[str, str]:
+        with self.stage_lock:
+            return dict(self.worker_stages)
+
+    def status_snapshot(self) -> dict[tuple[str, str], str]:
+        with self.unit_status_lock:
+            return dict(self.unit_statuses)
+
+
+def _harvest_finished_futures(futures: dict[Future[None], str]) -> None:
+    """Drop every finished future, re-raising the first failure."""
+
+    for future in [future for future in futures if future.done()]:
+        future.result()
+        del futures[future]
+
+
+@dataclass
+class _WarmWorkerPool:
+    """Warm-worker state machine for one materialization run.
+
+    A worker starts one execution unit as soon as the provider reports it
+    healthy, then takes the next unit from its own group's queue each time the
+    running unit succeeds. It stops once that queue is empty.
+    """
+
+    args: argparse.Namespace
+    pool: ThreadPoolExecutor
+    console: Console
+    channels: _DispatchChannels
+    assignment_queues: list[deque[WorkerAssignment]]
+    worker_group_index: dict[str, int]
+    worker_args: dict[str, argparse.Namespace]
+    pending: set[str]
+    futures: dict[Future[None], str] = field(default_factory=dict)
+    current_assignments: dict[str, WorkerAssignment] = field(default_factory=dict)
+    completed_unit_ids: set[str] = field(default_factory=set)
+
+    @property
+    def busy(self) -> bool:
+        """Report whether any worker still has a bootstrap, unit, or stop outstanding."""
+
+        return bool(
+            self.pending
+            or self.futures
+            or self.current_assignments
+            or any(self.assignment_queues)
+        )
+
+    def advance_completed_units(self) -> None:
+        """Refill or stop every warm worker whose unit is no longer running.
+
+        Raises PreparationError when a unit reports failure or when a worker
+        stops before its unit reports success.
+        """
+
+        inflight_ids = set(self.futures.values())
+        status_snapshot = self.channels.status_snapshot()
+        self.channels.status_changed.clear()
+        stage_snapshot = self.channels.stage_snapshot()
+        for instance_id, assignment in list(self.current_assignments.items()):
+            if instance_id in inflight_ids:
+                continue
+            unit_id = assignment.rows[0]["unit_id"]
+            status = status_snapshot.get((instance_id, unit_id))
+            if status == "succeeded":
+                self.completed_unit_ids.add(unit_id)
+                del self.current_assignments[instance_id]
+                self._dispatch_next_unit(instance_id)
+            elif status is not None and status.startswith("failed"):
+                self.channels.set_stage(instance_id, "failed")
+                raise PreparationError(
+                    f"Execution unit {unit_id} failed on {instance_id}: {status}"
+                )
+            elif stage_snapshot.get(instance_id) in {"stopping", "stopped"}:
+                raise PreparationError(
+                    f"Worker {instance_id} stopped before execution unit "
+                    f"{unit_id} reported success"
+                )
+
+    def start_ready_workers(self) -> None:
+        """Bootstrap and dispatch every worker the provider now reports healthy."""
+
+        newly_ready = (
+            _ready_worker_ids(self.args, sorted(self.pending)) if self.pending else set()
+        )
+        for instance_id in sorted(newly_ready):
+            queue = self._queue_for(instance_id)
+            if not queue:
+                self.pending.remove(instance_id)
+                self._stop_worker(instance_id)
+                continue
+            assignment = queue.popleft()
+            self.pending.remove(instance_id)
+            self.current_assignments[instance_id] = assignment
+            self.channels.set_stage(instance_id, "queued")
+            self._submit(
+                instance_id,
+                _bootstrap_and_dispatch_worker,
+                self.args,
+                instance_id,
+                assignment,
+                self.channels.worker_stages,
+                self.channels.stage_lock,
+                self.console,
+            )
+
+    def _dispatch_next_unit(self, instance_id: str) -> None:
+        """Give one worker its next unit, or stop the worker when none remain."""
+
+        queue = self._queue_for(instance_id)
+        if not queue:
+            self._stop_worker(instance_id)
+            return
+        assignment = queue.popleft()
+        self.current_assignments[instance_id] = assignment
+        self._submit(
+            instance_id,
+            _dispatch_assignment_to_worker,
+            instance_id,
+            assignment,
+            self.channels.worker_stages,
+            self.channels.stage_lock,
+            self.console,
+        )
+
+    def _stop_worker(self, instance_id: str) -> None:
+        self._submit(
+            instance_id,
+            _stop_worker_after_work,
+            instance_id,
+            self.worker_args[instance_id],
+            self.channels.worker_stages,
+            self.channels.stage_lock,
+            self.console,
+        )
+
+    def _queue_for(self, instance_id: str) -> deque[WorkerAssignment]:
+        """Return the queue of units this worker's instance type can run."""
+
+        return self.assignment_queues[self.worker_group_index[instance_id]]
+
+    def _submit(
+        self, instance_id: str, function: Callable[..., None], *call_args: Any
+    ) -> None:
+        self.futures[self.pool.submit(function, *call_args)] = instance_id
+
+
+def _await_dispatch_progress(
+    args: argparse.Namespace,
+    workers: _WarmWorkerPool,
+    resume_futures: dict[Future[None], str],
+) -> None:
+    """Block until the next event that can change the dispatch loop's decisions."""
+
+    if workers.pending:
+        # Readiness is visible only through the provider, so this wait must time
+        # out on the readiness poll interval.
+        active_futures = [*workers.futures, *resume_futures]
+        if active_futures:
+            wait(
+                active_futures,
+                timeout=args.readiness_poll_seconds,
+                return_when=FIRST_COMPLETED,
+            )
+        else:
+            time.sleep(args.readiness_poll_seconds)
+    elif workers.futures:
+        wait(workers.futures, return_when=FIRST_COMPLETED)
+    elif resume_futures:
+        wait(resume_futures, return_when=FIRST_COMPLETED)
+    elif workers.current_assignments:
+        workers.channels.status_changed.wait(args.readiness_poll_seconds)
+
+
+def _execute_materialization_groups(
+    args: argparse.Namespace,
+    groups: Sequence[MaterializationGroup],
+    selected: Sequence[dict[str, str]],
+    status_run_id: str,
+) -> None:
+    """Dispatch each worker as soon as it becomes ready."""
+
+    prepared: list[tuple[MaterializationGroup, list[str]]] = []
+    all_worker_ids: list[str] = []
+    deferred_resumes: list[tuple[argparse.Namespace, list[str]]] = []
+    assignment_queues = [deque(_stage_worker_assignments(group)) for group in groups]
+    try:
+        for group_index, group in enumerate(groups):
+            group_resume_ids: list[str] = []
+            worker_ids = _prepare_workers(
+                group.args,
+                group.worker_count,
+                owned_instance_ids=all_worker_ids,
+                wait_for_ready=False,
+                delay_after_last_batch=group_index < len(groups) - 1,
+                deferred_resume_ids=group_resume_ids,
+            )
+            prepared.append((group, worker_ids))
+            all_worker_ids.extend(worker_ids)
+            if group_resume_ids:
+                deferred_resumes.append((group.args, group_resume_ids))
+
+        channels = _DispatchChannels(
+            worker_stages={instance_id: "waiting" for instance_id in all_worker_ids}
+        )
+        lifecycle_console = Console(stderr=True, highlight=False)
+        worker_group_index = {
+            instance_id: group_index
+            for group_index, (_, worker_ids) in enumerate(prepared)
+            for instance_id in worker_ids
+        }
+        worker_args = {
+            instance_id: group.args
+            for group, worker_ids in prepared
+            for instance_id in worker_ids
+        }
+        resume_pool = ThreadPoolExecutor(max_workers=1)
+        resume_futures = (
+            {
+                resume_pool.submit(
+                    _resume_worker_groups_in_batches,
+                    deferred_resumes,
+                    console=lifecycle_console,
+                ): "resume workers"
+            }
+            if deferred_resumes
+            else {}
+        )
+        monitor_pool = ThreadPoolExecutor(max_workers=1)
+        monitor = monitor_pool.submit(
+            _wait_for_workers_to_stop,
+            args,
+            all_worker_ids,
+            len(selected),
+            status_run_id,
+            worker_stages=channels.worker_stages,
+            stage_lock=channels.stage_lock,
+            all_dispatched=channels.all_dispatched,
+            abort=channels.abort,
+            console=lifecycle_console,
+            worker_args=worker_args,
+            unit_statuses=channels.unit_statuses,
+            unit_status_lock=channels.unit_status_lock,
+            status_changed=channels.status_changed,
+        )
+        try:
+            bootstrap_workers = min(args.bootstrap_parallelism, len(all_worker_ids))
+            with ThreadPoolExecutor(max_workers=bootstrap_workers) as pool:
+                workers = _WarmWorkerPool(
+                    args=args,
+                    pool=pool,
+                    console=lifecycle_console,
+                    channels=channels,
+                    assignment_queues=assignment_queues,
+                    worker_group_index=worker_group_index,
+                    worker_args=worker_args,
+                    pending=set(all_worker_ids),
+                )
+                while workers.busy or resume_futures:
+                    _harvest_finished_futures(resume_futures)
+                    _harvest_finished_futures(workers.futures)
+                    workers.advance_completed_units()
+                    workers.start_ready_workers()
+                    _await_dispatch_progress(args, workers, resume_futures)
+
+            channels.all_dispatched.set()
+            monitor.result()
+        except BaseException:
+            channels.abort.set()
+            raise
+        finally:
+            monitor_pool.shutdown(wait=True)
+            resume_pool.shutdown(wait=True)
+
+        if any(assignment_queues):
+            raise PreparationError("Not every execution-unit assignment was dispatched")
+        if len(workers.completed_unit_ids) != len(selected):
+            raise PreparationError(
+                f"Only {len(workers.completed_unit_ids):,}/{len(selected):,} execution units "
+                "reported success"
+            )
+        _verify_materialized_units(args, selected)
+    except BaseException:
+        for group, worker_ids in prepared:
+            _pause_workers_after_failure(group.args, worker_ids)
+        raise
+
+
+def main() -> None:
+    parser = build_parser()
+    try:
+        args = parser.parse_args()
+        args.region = normalize_region(args.region)
+        if args.preflight and not args.execute:
+            raise PreparationError("--preflight requires --execute")
+        build = args.build.resolve()
+        rows = _load_execution_units(build)
+        if args.list_categories is not None:
+            _print_categories(rows, args.list_categories)
+            return
+
+        label, selected = _select_units(args, rows)
+        worker_groups = _planned_worker_groups(args, selected)
+        status_run_id = f"{int(time.time())}-{uuid.uuid4().hex[:12]}" if args.execute else None
+        if args.execute:
+            if shutil.which("pmr") is None:
+                raise PreparationError(
+                    "pmr is unavailable; run materialize.py with uv so its inline dependencies are installed"
+                )
+            for required_path in (
+                WORKER_STORAGE_SCRIPT,
+                RESHARD_MODULE,
+                DOCUMENT_SELECTION_MODULE,
+            ):
+                if required_path.is_symlink() or not required_path.is_file():
+                    raise PreparationError(f"Required worker runtime file is missing or unsafe: {required_path}")
+            if args.profile:
+                os.environ["AWS_PROFILE"] = args.profile
+            if args.preflight:
+                _run_selected_preflight(args, build, selected)
+            preflight_created_at = _require_preflight(build, selected)
+            print(f"preflight=passed created_at={preflight_created_at}")
+
+        worker_counts = _worker_counts_for_groups(worker_groups, args.parallelism)
+        materialization_groups: list[MaterializationGroup] = []
+        for group, worker_count in zip(worker_groups, worker_counts):
+            group_args = copy.copy(args)
+            group_args.instance_type = group.instance_type
+            group_args.storage_layout = group.storage_layout
+            group_rows = list(group.rows)
+            group_label = label if len(worker_groups) == 1 else f"{label}-{group.instance_type}"
+            script_dir = _stage_selection(build, group_label, group_rows, status_run_id)
+            lifecycle_commands = _dry_run_lifecycle_commands(
+                group_args,
+                group_rows,
+                script_dir,
+                worker_count,
+            )
+            _print_dispatch(
+                group_label,
+                group_rows,
+                script_dir,
+                lifecycle_commands,
+                worker_count,
+                args.execute,
+                cluster=args.cluster,
+                project=args.project,
+                region=args.region,
+                instance_type=group.instance_type,
+                storage_layout=group.storage_layout,
+            )
+            materialization_groups.append(
+                MaterializationGroup(
+                    args=group_args,
+                    rows=tuple(group_rows),
+                    script_dir=script_dir,
+                    worker_count=worker_count,
+                )
+            )
+
+        if args.execute:
+            assert status_run_id is not None
+            _execute_materialization_groups(
+                args,
+                materialization_groups,
+                selected,
+                status_run_id,
+            )
+    except PreparationError as exc:
+        parser.exit(2, f"error: {exc}\n")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/dolma3p5_resharding/output_report.py b/scripts/dolma3p5_resharding/output_report.py
new file mode 100644
index 00000000..9e28e01e
--- /dev/null
+++ b/scripts/dolma3p5_resharding/output_report.py
@@ -0,0 +1,502 @@
+"""Hierarchical source, target, and materialized-output reporting."""
+
+from __future__ import annotations
+
+import html
+from collections import Counter, defaultdict
+from typing import Any, Sequence
+
+# One light/dark value pair per report color role. Report stylesheets name the
+# roles they use instead of repeating hex literals.
+REPORT_PALETTE: dict[str, tuple[str, str]] = {
+    "muted": ("#536965", "#a7bbb7"),
+    "surface": ("#edf6f4", "#142420"),
+    "surface-raised": ("#e4f0ed", "#1a2d29"),
+    "surface-hover": ("#eaf3f1", "#172522"),
+    "surface-selected": ("#dceeea", "#1b312d"),
+    "code": ("#e2efec", "#1b312d"),
+    "track": ("#d2e1de", "#2a403c"),
+    "line": ("#cbdad7", "#2a403c"),
+    "accent": ("#14786f", "#5cc8bb"),
+    "accent-output": ("#218f84", "#5cc8bb"),
+    "accent-link": ("#126a63", "#74d7cb"),
+    "neutral-series": ("#71817e", "#91a29f"),
+    "neutral-target": ("#7b8d89", "#91a29f"),
+    "warning": ("#a35f16", "#e5a456"),
+    "warning-selection": ("#d28a32", "#e5a456"),
+    "warning-strong": ("#bd711f", "#e5a456"),
+    "shell": ("#f4f8f7", "#0e1715"),
+    "shell-tab": ("#e1ece9", "#172522"),
+}
+
+
+def report_root_style(variables: Sequence[tuple[str, str]]) -> str:
+    """Render one report's light and dark :root custom-property blocks.
+
+    Each pair is a CSS custom-property name and the REPORT_PALETTE role
+    supplying its light and dark values.
+    """
+
+    def block(index: int) -> str:
+        return ";".join(
+            f"--{name}:{REPORT_PALETTE[role][index]}" for name, role in variables
+        )
+
+    return (
+        f":root{{color-scheme:light dark;{block(0)}}}\n"
+        f"@media(prefers-color-scheme:dark){{:root{{{block(1)}}}}}"
+    )
+
+
+def _human_count(value: int) -> str:
+    for scale, suffix in (
+        (1_000_000_000_000, "T"),
+        (1_000_000_000, "B"),
+        (1_000_000, "M"),
+        (1_000, "K"),
+    ):
+        if abs(value) >= scale:
+            return f"{value / scale:.3g}{suffix}"
+    return f"{value:,}"
+
+
+def _multiplier(value: float) -> str:
+    if value >= 100:
+        return f"{value:,.0f}×"
+    if value >= 10:
+        return f"{value:.1f}×"
+    return f"{value:.2f}×"
+
+
+def _status(statuses_by_leaf: dict[str, list[str]], leaves: Sequence[str]) -> str:
+    statuses = [status for leaf in leaves for status in statuses_by_leaf.get(leaf, [])]
+    if not statuses:
+        return "dropped"
+    return "passed" if all(status == "passed" for status in statuses) else "outside_tolerance"
+
+
+def _node(
+    name: str,
+    source: int,
+    target: int,
+    actual: int | None,
+    status: str,
+    **extra: Any,
+) -> dict[str, Any]:
+    return {
+        "name": name,
+        "source": source,
+        "target": target,
+        "actual": actual,
+        "status": status,
+        **extra,
+    }
+
+
+def _csv_row(
+    *,
+    level: str,
+    source_family: str,
+    source: int,
+    target: int,
+    actual: int | None,
+    status: str,
+    subcategory_name: str = "",
+    category_name: str = "",
+    lower_group: str = "",
+    leaf_id: str = "",
+) -> dict[str, Any]:
+    return {
+        "level": level,
+        "source_family": source_family,
+        "subcategory_name": subcategory_name,
+        "category_name": category_name,
+        "lower_group": lower_group,
+        "leaf_id": leaf_id,
+        "source_uint32_values": source,
+        "target_uint32_values": target,
+        "actual_uint32_values": "" if actual is None else actual,
+        "actual_minus_target": "" if actual is None else actual - target,
+        "planned_sampling_ratio": "" if source == 0 else f"{target / source:.12g}",
+        "realized_sampling_ratio": ("" if actual is None or source == 0 else f"{actual / source:.12g}"),
+        "status": status,
+    }
+
+
+def _build_hierarchy(
+    inventory_details: dict[str, Any],
+    validation_rows: Sequence[dict[str, Any]],
+) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
+    actual_by_leaf: dict[str, int] = defaultdict(int)
+    unit_count_by_leaf: Counter[str] = Counter()
+    shard_count_by_leaf: Counter[str] = Counter()
+    statuses_by_leaf: dict[str, list[str]] = defaultdict(list)
+    for row in validation_rows:
+        leaf_id = str(row["leaf_id"])
+        actual_by_leaf[leaf_id] += int(row["actual_uint32_values"])
+        unit_count_by_leaf[leaf_id] += 1
+        shard_count_by_leaf[leaf_id] += int(row["npy_count"])
+        statuses_by_leaf[leaf_id].append(str(row["status"]))
+    selected_leaf_ids = set(actual_by_leaf)
+    selected_mix_names = {str(row["mix_name"]) for row in validation_rows}
+    flat_rows: list[dict[str, Any]] = []
+    subcategories: list[dict[str, Any]] = []
+
+    for source_row in inventory_details.get("sources", []):
+        mix_name = str(source_row["mix_name"])
+        if mix_name not in selected_mix_names:
+            continue
+        family_name = str(source_row["source_family"])
+        subcategory_name = str(source_row["subcategory_name"])
+        categories: list[dict[str, Any]] = []
+        for category_row in source_row.get("categories", []):
+            leaf_id = str(category_row["leaf_id"])
+            source = int(category_row["source_uint32_values"])
+            target = int(category_row["target_uint32_values"])
+            if leaf_id not in selected_leaf_ids and target > 0:
+                continue
+            actual = actual_by_leaf.get(leaf_id, 0)
+            status = _status(statuses_by_leaf, [leaf_id]) if leaf_id in selected_leaf_ids else "dropped"
+            lower_groups: list[dict[str, Any]] = []
+            for lower_row in category_row.get("lower_groups", []):
+                lower_source = int(lower_row["source_uint32_values"])
+                lower_target = int(lower_row["implied_target_uint32_values"])
+                lower_actual = 0 if status == "dropped" else None
+                lower = _node(
+                    str(lower_row["lower_group"]),
+                    lower_source,
+                    lower_target,
+                    lower_actual,
+                    "dropped" if lower_actual == 0 else "planned_only",
+                    npy_count=int(lower_row["unique_npy_count"]),
+                )
+                lower_groups.append(lower)
+                flat_rows.append(
+                    _csv_row(
+                        level="lower_group",
+                        source_family=family_name,
+                        subcategory_name=subcategory_name,
+                        category_name=str(category_row["category_name"]),
+                        lower_group=lower["name"],
+                        leaf_id=leaf_id,
+                        source=lower_source,
+                        target=lower_target,
+                        actual=lower_actual,
+                        status=lower["status"],
+                    )
+                )
+            category = _node(
+                str(category_row["category_name"]),
+                source,
+                target,
+                actual,
+                status,
+                leaf_id=leaf_id,
+                unit_count=unit_count_by_leaf[leaf_id],
+                output_shards=shard_count_by_leaf[leaf_id],
+                lower_groups=sorted(
+                    lower_groups,
+                    key=lambda row: (row["target"], row["source"], row["name"]),
+                    reverse=True,
+                ),
+            )
+            categories.append(category)
+            flat_rows.append(
+                _csv_row(
+                    level="category",
+                    source_family=family_name,
+                    subcategory_name=subcategory_name,
+                    category_name=category["name"],
+                    leaf_id=leaf_id,
+                    source=source,
+                    target=target,
+                    actual=actual,
+                    status=status,
+                )
+            )
+        if not categories:
+            continue
+        source = sum(row["source"] for row in categories)
+        target = sum(row["target"] for row in categories)
+        actual = sum(row["actual"] for row in categories)
+        leaves = [row["leaf_id"] for row in categories]
+        status = _status(statuses_by_leaf, leaves)
+        subcategory = _node(
+            subcategory_name,
+            source,
+            target,
+            actual,
+            status,
+            source_family=family_name,
+            mix_name=mix_name,
+            leaf_ids=leaves,
+            categories=sorted(
+                categories,
+                key=lambda row: (row["target"], row["source"], row["name"]),
+                reverse=True,
+            ),
+        )
+        subcategories.append(subcategory)
+        flat_rows.append(
+            _csv_row(
+                level="subcategory",
+                source_family=family_name,
+                subcategory_name=subcategory_name,
+                source=source,
+                target=target,
+                actual=actual,
+                status=status,
+            )
+        )
+
+    by_family: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for subcategory in subcategories:
+        by_family[subcategory["source_family"]].append(subcategory)
+    families: list[dict[str, Any]] = []
+    for family_name, children in by_family.items():
+        source = sum(row["source"] for row in children)
+        target = sum(row["target"] for row in children)
+        actual = sum(row["actual"] for row in children)
+        leaves = [leaf for row in children for leaf in row["leaf_ids"]]
+        status = _status(statuses_by_leaf, leaves)
+        family = _node(
+            family_name,
+            source,
+            target,
+            actual,
+            status,
+            subcategories=sorted(
+                children,
+                key=lambda row: (row["target"], row["source"], row["name"]),
+                reverse=True,
+            ),
+        )
+        families.append(family)
+        flat_rows.append(
+            _csv_row(
+                level="source_family",
+                source_family=family_name,
+                source=source,
+                target=target,
+                actual=actual,
+                status=status,
+            )
+        )
+    families.sort(
+        key=lambda row: (row["target"], row["source"], row["name"]),
+        reverse=True,
+    )
+    flat_rows.sort(
+        key=lambda row: (
+            row["source_family"],
+            row["subcategory_name"],
+            row["category_name"],
+            row["lower_group"],
+            row["level"],
+        )
+    )
+    return families, flat_rows
+
+
+def _metric(label: str, value: str, warning: bool = False) -> str:
+    warning_class = " validation-warning" if warning else ""
+    return (
+        f'<span class="validation-metric{warning_class}">'
+        f'<span class="validation-label">{html.escape(label)}</span>'
+        f'<span class="validation-value">{html.escape(value)}</span></span>'
+    )
+
+
+def _sampling_label(source: int, target: int, actual: int | None) -> str:
+    if source <= 0:
+        return "—"
+    planned = _multiplier(target / source)
+    if actual is None:
+        return planned
+    return f"{planned} planned · {_multiplier(actual / source)} actual"
+
+
+def _delta_label(target: int, actual: int) -> str:
+    delta = actual - target
+    sign = "+" if delta > 0 else ""
+    percent = 100 * delta / target if target else 0.0
+    return f"{sign}{_human_count(delta)} · {percent:+.5f}%"
+
+
+def _bars(target: int, actual: int | None, maximum: int) -> str:
+    target_width = 100 * target / maximum if maximum else 0.0
+    actual_width = 100 * actual / maximum if actual is not None and maximum else 0.0
+    actual_bar = (
+        '<span class="actual-track"><span class="actual-fill" ' f'style="width:{actual_width:.8f}%"></span></span>'
+        if actual is not None
+        else ""
+    )
+    return (
+        '<span class="target-actual-bars" aria-hidden="true">'
+        '<span class="target-track"><span class="target-fill" '
+        f'style="width:{target_width:.8f}%"></span></span>'
+        f"{actual_bar}</span>"
+    )
+
+
+def _values(node: dict[str, Any], maximum: int) -> str:
+    actual = node["actual"]
+    return (
+        _metric("Source", f"{_human_count(node['source'])} tokens")
+        + _metric("Target", f"{_human_count(node['target'])} tokens")
+        + _metric(
+            "Actual",
+            "—" if actual is None else f"{_human_count(actual)} tokens",
+        )
+        + _metric(
+            "Sampling",
+            _sampling_label(node["source"], node["target"], actual),
+        )
+        + (
+            _metric(
+                "Actual − target",
+                _delta_label(node["target"], actual),
+                warning=node["status"] == "outside_tolerance",
+            )
+            if actual is not None
+            else _metric("Actual − target", "—")
+        )
+        + _bars(node["target"], actual, maximum)
+    )
+
+
+def _render_hierarchy(families: Sequence[dict[str, Any]]) -> str:
+    family_maximum = max(
+        (max(row["target"], row["actual"]) for row in families),
+        default=1,
+    )
+    family_sections: list[str] = []
+    for family in families:
+        subcategory_maximum = max(
+            (max(row["target"], row["actual"]) for row in family["subcategories"]),
+            default=1,
+        )
+        subcategory_sections: list[str] = []
+        for subcategory in family["subcategories"]:
+            category_maximum = max(
+                (max(row["target"], row["actual"]) for row in subcategory["categories"]),
+                default=1,
+            )
+            category_sections: list[str] = []
+            for category in subcategory["categories"]:
+                lower_maximum = max(
+                    (row["target"] for row in category["lower_groups"]),
+                    default=1,
+                )
+                lower_rows = "".join(
+                    '<div class="validation-row lower-group-row">'
+                    f'<span class="validation-name">{html.escape(lower["name"])}'
+                    f'<span class="validation-context">{lower["npy_count"]:,} NPY</span></span>'
+                    + _values(lower, lower_maximum)
+                    + "</div>"
+                    for lower in category["lower_groups"]
+                )
+                context = (
+                    "not selected by the sampling plan"
+                    if category["status"] == "dropped"
+                    else (
+                        f"{category['unit_count']:,} unit"
+                        f"{'s' if category['unit_count'] != 1 else ''} · "
+                        f"{category['output_shards']:,} output shard"
+                        f"{'s' if category['output_shards'] != 1 else ''}"
+                    )
+                )
+                category_sections.append(
+                    '<details class="validation-node category-node">'
+                    '<summary class="validation-row">'
+                    f'<span class="validation-name">{html.escape(category["name"])}'
+                    f'<span class="validation-context">{html.escape(context)}</span></span>'
+                    + _values(category, category_maximum)
+                    + "</summary>"
+                    + (
+                        '<div class="validation-children lower-groups">' + lower_rows + "</div>"
+                        if lower_rows
+                        else ""
+                    )
+                    + "</details>"
+                )
+            subcategory_sections.append(
+                '<details class="validation-node subcategory-node">'
+                '<summary class="validation-row">'
+                f'<span class="validation-name">{html.escape(subcategory["name"])}</span>'
+                + _values(subcategory, subcategory_maximum)
+                + '</summary><div class="validation-children">'
+                + "".join(category_sections)
+                + "</div></details>"
+            )
+        family_sections.append(
+            '<details class="validation-node family-node">'
+            '<summary class="validation-row">'
+            f'<span class="validation-name">{html.escape(family["name"])}</span>'
+            + _values(family, family_maximum)
+            + '</summary><div class="validation-children">'
+            + "".join(subcategory_sections)
+            + "</div></details>"
+        )
+    return "".join(family_sections)
+
+
+def _style() -> str:
+    return (
+        "\n<style>\n"
+        + report_root_style(
+            (
+                ("muted", "muted"),
+                ("surface", "surface"),
+                ("surface-2", "surface-raised"),
+                ("track", "track"),
+                ("target", "neutral-target"),
+                ("actual", "accent-output"),
+                ("warning", "warning-strong"),
+            )
+        )
+        + """
+*{box-sizing:border-box}body{max-width:1480px;margin:0 auto;padding:34px 26px 72px;background:Canvas;color:CanvasText;font:14px/1.42 system-ui,sans-serif}h1{margin:0 0 20px;font-size:28px;line-height:1.2}.aggregate-row{display:grid;grid-template-columns:repeat(5,minmax(150px,1fr));gap:12px 28px;margin-bottom:16px}.plot-legend{display:flex;gap:18px;margin:0 0 8px;color:var(--muted)}.plot-legend span::before{display:inline-block;width:18px;height:5px;margin-right:7px;border-radius:999px;content:"";vertical-align:middle}.target-key::before{background:var(--target)}.actual-key::before{background:var(--actual)}.provenance-note{max-width:940px;margin:0 0 18px;color:var(--muted)}.validation-hierarchy{display:grid;gap:7px}.validation-node{border:0}.validation-row{display:grid;grid-template-columns:minmax(250px,1.35fr) repeat(5,minmax(128px,1fr));gap:8px 22px;align-items:center;min-width:0;padding:12px 14px;list-style-position:inside}.validation-node>summary{cursor:pointer}.family-node>summary{border-radius:9px;background:var(--surface)}.family-node>summary:hover,.subcategory-node>summary:hover,.category-node>summary:hover{background:var(--surface-2)}.subcategory-node>summary{margin-top:6px;border-radius:8px;background:color-mix(in srgb,var(--surface) 62%,transparent)}.category-node>summary{border-radius:7px}.validation-name{min-width:0;overflow-wrap:anywhere;font-weight:600}.validation-context{display:block;margin:2px 0 0 18px;color:var(--muted);font-size:12px;font-weight:400}.validation-metric{min-width:0;font-variant-numeric:tabular-nums}.validation-label,.validation-value{display:block}.validation-label{color:var(--muted);font-size:12px}.validation-value{margin-top:2px}.validation-warning .validation-value{color:var(--warning)}.target-actual-bars{grid-column:2/-1;display:grid;gap:3px}.target-track,.actual-track{display:block;height:5px;border-radius:999px;background:var(--track);overflow:hidden}.target-fill,.actual-fill{display:block;height:100%;border-radius:inherit}.target-fill{background:var(--target)}.actual-fill{background:var(--actual)}.validation-children{margin-left:22px}.lower-groups{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:3px 22px}.lower-group-row{grid-template-columns:minmax(190px,1.15fr) repeat(5,minmax(105px,1fr));padding-top:9px;padding-bottom:9px}.lower-group-row .target-actual-bars{grid-column:2/-1}
+@media(max-width:1050px){.validation-row,.lower-group-row{grid-template-columns:minmax(220px,1fr) repeat(2,minmax(130px,1fr))}.validation-row>.validation-metric:nth-of-type(n+4){margin-top:5px}.target-actual-bars,.lower-group-row .target-actual-bars{grid-column:2/-1}.lower-groups{grid-template-columns:1fr}}
+@media(max-width:700px){body{padding:24px 14px 48px}.aggregate-row{grid-template-columns:repeat(2,minmax(0,1fr))}.validation-row,.lower-group-row{grid-template-columns:1fr 1fr;gap:8px 14px}.validation-name{grid-column:1/-1}.target-actual-bars,.lower-group-row .target-actual-bars{grid-column:1/-1}.validation-children{margin-left:8px}}
+</style>
+"""
+    )
+
+
+def render_output_validation_report(
+    inventory_details: dict[str, Any],
+    validation_rows: Sequence[dict[str, Any]],
+) -> tuple[list[dict[str, Any]], str]:
+    """Render materialized output totals against the inventory and sampling plan.
+
+    Returns the flat hierarchy rows for the CSV and the report HTML.
+    """
+
+    families, flat_rows = _build_hierarchy(inventory_details, validation_rows)
+    total_source = sum(row["source"] for row in families)
+    total_target = sum(row["target"] for row in families)
+    total_actual = sum(row["actual"] for row in families)
+    report = (
+        '<!doctype html><html><head><meta charset="utf-8">'
+        '<meta name="viewport" content="width=device-width,initial-scale=1">'
+        "<title>Dolma 3.5 source, target, and materialized output</title>" + _style() + "</head><body>"
+        "<h1>Dolma 3.5 Source, Target, and Materialized Output</h1>"
+        '<div class="aggregate-row">'
+        + _metric("Source", f"{_human_count(total_source)} tokens")
+        + _metric("Target", f"{_human_count(total_target)} tokens")
+        + _metric("Actual", f"{_human_count(total_actual)} tokens")
+        + _metric(
+            "Sampling",
+            _sampling_label(total_source, total_target, total_actual),
+        )
+        + _metric("Actual − target", _delta_label(total_target, total_actual))
+        + "</div>"
+        '<div class="plot-legend"><span class="target-key">Target</span>'
+        '<span class="actual-key">Actual</span></div>'
+        '<p class="provenance-note">Actual output is exact through the sampling tier. '
+        "Lower-group rows retain the source and target plan because merged output does not retain "
+        "per-input provenance.</p>"
+        '<main class="validation-hierarchy">' + _render_hierarchy(families) + "</main></body></html>\n"
+    )
+    return flat_rows, report
diff --git a/scripts/dolma3p5_resharding/plan.py b/scripts/dolma3p5_resharding/plan.py
new file mode 100644
index 00000000..f653e822
--- /dev/null
+++ b/scripts/dolma3p5_resharding/plan.py
@@ -0,0 +1,120 @@
+"""Build the complete Dolma 3.5 sampling and materialization plan."""
+
+from __future__ import annotations
+
+import argparse
+import shutil
+from pathlib import Path
+
+try:
+    from .workflow import (
+        DEFAULT_BUILD_PATH,
+        DEFAULT_REGION,
+        PreparationError,
+        collect_inventory,
+        plan_build,
+        propose_configs,
+    )
+except ImportError:
+    from workflow import (
+        DEFAULT_BUILD_PATH,
+        DEFAULT_REGION,
+        PreparationError,
+        collect_inventory,
+        plan_build,
+        propose_configs,
+    )
+
+REPOSITORY_ROOT = Path(__file__).resolve().parents[2]
+CONFIG_ROOT = REPOSITORY_ROOT / "configs/dolma3p5-resharding/14t"
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description=__doc__,
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
+    )
+    parser.add_argument(
+        "--mix",
+        type=Path,
+        default=CONFIG_ROOT / "inputs/mix-v2.yaml",
+        help="authoritative mix YAML; inputs/dolma3p5-14t-optimal-mix.yaml is the superseded v1",
+    )
+    parser.add_argument(
+        "--catalog",
+        type=Path,
+        default=CONFIG_ROOT / "inputs/dolma3p5-reference-all-npy-s3-paths.csv",
+        help="reference superset of S3 NPY paths",
+    )
+    parser.add_argument(
+        "--settings",
+        type=Path,
+        default=CONFIG_ROOT / "settings.yaml",
+        help="preparation settings",
+    )
+    parser.add_argument(
+        "--output",
+        type=Path,
+        default=DEFAULT_BUILD_PATH,
+        help="local preparation build directory",
+    )
+    parser.add_argument("--profile", help="optional AWS profile")
+    parser.add_argument(
+        "--region",
+        default=DEFAULT_REGION,
+        help="AWS region; override when the source store is in another region",
+    )
+    parser.add_argument(
+        "--max-workers",
+        type=int,
+        help="override concurrent read-only exact-object requests",
+    )
+    parser.add_argument(
+        "--destination-root",
+        required=True,
+        help="new object-store prefix below which every execution unit gets a unique destination",
+    )
+    parser.add_argument(
+        "--local-temp-root",
+        type=Path,
+        required=True,
+        help="absolute worker-local directory used while materializing one execution unit",
+    )
+    parser.add_argument(
+        "--max-unit-working-bytes",
+        type=int,
+        required=True,
+        help="maximum estimated local working set for one execution unit",
+    )
+    return parser
+
+
+def main() -> None:
+    parser = build_parser()
+    try:
+        args = parser.parse_args()
+        if shutil.which("s5cmd") is None:
+            raise PreparationError("s5cmd is required for planning and was not found on PATH")
+        plan_build(args)
+        collect_inventory(
+            argparse.Namespace(
+                build=args.output,
+                profile=args.profile,
+                region=args.region,
+                max_workers=args.max_workers,
+            )
+        )
+        propose_configs(
+            argparse.Namespace(
+                build=args.output,
+                destination_root=args.destination_root,
+                local_temp_root=args.local_temp_root,
+                max_unit_working_bytes=args.max_unit_working_bytes,
+            )
+        )
+    except PreparationError as exc:
+        parser.exit(2, f"error: {exc}\n")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/dolma3p5_resharding/preflight.py b/scripts/dolma3p5_resharding/preflight.py
new file mode 100644
index 00000000..d6930863
--- /dev/null
+++ b/scripts/dolma3p5_resharding/preflight.py
@@ -0,0 +1,55 @@
+"""Run read-only source-drift and destination checks before materialization."""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+
+from workflow import DEFAULT_BUILD_PATH, DEFAULT_REGION, PreparationError, preflight_build
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description=__doc__,
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
+    )
+    parser.add_argument(
+        "--build",
+        type=Path,
+        default=DEFAULT_BUILD_PATH,
+        help="local preparation build directory",
+    )
+    parser.add_argument("--profile", help="optional AWS profile")
+    parser.add_argument(
+        "--region",
+        default=DEFAULT_REGION,
+        help="AWS region; override when the source or destination store is in another region",
+    )
+    parser.add_argument("--max-workers", type=int)
+    selection = parser.add_mutually_exclusive_group()
+    selection.add_argument(
+        "--all",
+        action="store_true",
+        help="check every execution unit; this is also the default selection",
+    )
+    selection.add_argument(
+        "--category",
+        help="check one exact mix name, leaf ID, or MIX_NAME::CATEGORY_NAME selector",
+    )
+    selection.add_argument(
+        "--unit",
+        help="check one exact execution-unit ID",
+    )
+    return parser
+
+
+def main() -> None:
+    parser = build_parser()
+    try:
+        preflight_build(parser.parse_args())
+    except PreparationError as exc:
+        parser.exit(2, f"error: {exc}\n")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/dolma3p5_resharding/setup_worker_storage.sh b/scripts/dolma3p5_resharding/setup_worker_storage.sh
new file mode 100755
index 00000000..8e6e5526
--- /dev/null
+++ b/scripts/dolma3p5_resharding/setup_worker_storage.sh
@@ -0,0 +1,264 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+usage() {
+  cat <<'EOF'
+Usage: setup_worker_storage.sh --check|--apply [--layout auto|single|raid0]
+
+Prepare EC2 NVMe instance-store devices for Dolma resharding.
+
+  --check  Print detected devices and the direct-mount/RAID0 plan. No writes.
+  --apply  Format only unused EC2 instance-store devices and mount /mnt/dolma.
+  --layout Storage layout (default: auto). Auto selects a single device when
+           only one exists and requires an explicit choice for multiple devices.
+
+Environment overrides:
+  DOLMA_WORK_MOUNT          Mount point (default: /mnt/dolma)
+  DOLMA_MIN_AVAILABLE_BYTES Minimum required free bytes (default: 1650000000000)
+  DOLMA_STORAGE_LAYOUT      Default value for --layout
+EOF
+}
+
+mode=""
+layout="${DOLMA_STORAGE_LAYOUT:-auto}"
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --check|--apply)
+      if [[ -n "$mode" ]]; then
+        echo "Specify exactly one of --check or --apply" >&2
+        exit 2
+      fi
+      mode="$1"
+      ;;
+    --layout)
+      if [[ $# -lt 2 ]]; then
+        echo "--layout requires auto, single, or raid0" >&2
+        exit 2
+      fi
+      layout="$2"
+      shift
+      ;;
+    --layout=*) layout="${1#*=}" ;;
+    --help|-h) usage; exit 0 ;;
+    *) usage >&2; exit 2 ;;
+  esac
+  shift
+done
+
+if [[ -z "$mode" ]]; then
+  usage >&2
+  exit 2
+fi
+case "$layout" in
+  auto|single|raid0) ;;
+  *) echo "--layout must be auto, single, or raid0" >&2; exit 2 ;;
+esac
+
+mount_point="${DOLMA_WORK_MOUNT:-/mnt/dolma}"
+minimum_available_bytes="${DOLMA_MIN_AVAILABLE_BYTES:-1650000000000}"
+marker_path="$mount_point/.dolma3p5-instance-store"
+
+if ! [[ "$minimum_available_bytes" =~ ^[1-9][0-9]*$ ]]; then
+  echo "DOLMA_MIN_AVAILABLE_BYTES must be a positive integer" >&2
+  exit 2
+fi
+
+instance_devices=()
+while read -r device device_type; do
+  [[ "$device_type" == "disk" ]] || continue
+  model=$(lsblk -dno MODEL "$device" | sed 's/^[[:space:]]*//;s/[[:space:]]*$//')
+  if [[ "$model" == *"EC2 NVMe Instance Storage"* ]]; then
+    instance_devices+=("$device")
+  fi
+done < <(lsblk -dpno NAME,TYPE)
+
+if [[ ${#instance_devices[@]} -eq 0 ]]; then
+  echo "No EC2 NVMe instance-store devices were detected; refusing to continue" >&2
+  exit 1
+fi
+
+echo "Detected ${#instance_devices[@]} EC2 instance-store device(s):"
+instance_device_sizes=()
+total_bytes=0
+largest_device_index=0
+largest_device_bytes=0
+for device_index in "${!instance_devices[@]}"; do
+  device="${instance_devices[$device_index]}"
+  size_bytes=$(lsblk -bdno SIZE "$device")
+  instance_device_sizes+=("$size_bytes")
+  total_bytes=$((total_bytes + size_bytes))
+  if (( size_bytes > largest_device_bytes )); then
+    largest_device_index=$device_index
+    largest_device_bytes=$size_bytes
+  fi
+done
+
+# Add two percent for filesystem overhead when comparing raw device capacity
+# with the requested usable capacity.
+minimum_direct_device_bytes=$((minimum_available_bytes + minimum_available_bytes / 50))
+layout_decision_required=0
+case "$layout" in
+  single)
+    selected_devices=("${instance_devices[$largest_device_index]}")
+    selected_raw_bytes=$largest_device_bytes
+    ;;
+  raid0)
+    if [[ ${#instance_devices[@]} -lt 2 ]]; then
+      echo "RAID0 requires at least two EC2 instance-store devices" >&2
+      exit 1
+    fi
+    selected_devices=("${instance_devices[@]}")
+    selected_raw_bytes=$total_bytes
+    ;;
+  auto)
+    if [[ ${#instance_devices[@]} -eq 1 ]]; then
+      layout="single"
+      selected_devices=("${instance_devices[0]}")
+      selected_raw_bytes=$largest_device_bytes
+    else
+      layout_decision_required=1
+      selected_devices=("${instance_devices[@]}")
+      selected_raw_bytes=$total_bytes
+    fi
+    ;;
+esac
+
+if (( layout_decision_required )); then
+  if (( total_bytes < minimum_direct_device_bytes )); then
+    echo "Combined instance-store capacity is too small for $minimum_available_bytes usable bytes" >&2
+    exit 1
+  fi
+elif (( selected_raw_bytes < minimum_direct_device_bytes )); then
+  echo "The selected $layout layout is too small for $minimum_available_bytes usable bytes" >&2
+  exit 1
+fi
+
+for device_index in "${!instance_devices[@]}"; do
+  device="${instance_devices[$device_index]}"
+  model=$(lsblk -dno MODEL "$device" | sed 's/^[[:space:]]*//;s/[[:space:]]*$//')
+  if (( layout_decision_required )); then
+    role="candidate"
+  else
+    role="unused"
+    for selected_device in "${selected_devices[@]}"; do
+      if [[ "$device" == "$selected_device" ]]; then
+        role="selected"
+        break
+      fi
+    done
+  fi
+  printf '  %s  %s bytes  %s  %s\n' \
+    "$device" "${instance_device_sizes[$device_index]}" "$role" "$model"
+done
+
+hazards=0
+for device in "${selected_devices[@]}"; do
+  if lsblk -nrpo NAME,TYPE "$device" | tail -n +2 | grep -q .; then
+    echo "    BLOCKED $device: device has child partitions or mappings" >&2
+    hazards=1
+  fi
+  if lsblk -nrpo MOUNTPOINT "$device" | grep -q '[^[:space:]]'; then
+    echo "    BLOCKED $device: device or a child is already mounted" >&2
+    hazards=1
+  fi
+  if sudo wipefs -n "$device" | grep -q .; then
+    echo "    BLOCKED $device: device already contains a filesystem or RAID signature" >&2
+    hazards=1
+  fi
+done
+
+if findmnt -rn "$mount_point" >/dev/null 2>&1; then
+  mounted_source=$(findmnt -rn -o SOURCE "$mount_point")
+  echo "$mount_point is already mounted from $mounted_source"
+  if [[ ! -f "$marker_path" ]]; then
+    echo "BLOCKED: existing mount is not marked as managed by this script" >&2
+    exit 1
+  fi
+  available_bytes=$(df -B1 --output=avail "$mount_point" | tail -n 1 | tr -d ' ')
+  if (( available_bytes < minimum_available_bytes )); then
+    echo "BLOCKED: only $available_bytes bytes are available at $mount_point" >&2
+    exit 1
+  fi
+  echo "Storage is already prepared with $available_bytes available bytes"
+  exit 0
+fi
+
+if (( layout_decision_required )); then
+  echo "Layout decision required for this multi-device worker:"
+  if (( largest_device_bytes >= minimum_direct_device_bytes )); then
+    echo "  --layout single: use ${instance_devices[$largest_device_index]} ($largest_device_bytes bytes raw)"
+  else
+    echo "  --layout single: insufficient capacity ($largest_device_bytes bytes raw)"
+  fi
+  echo "  --layout raid0: stripe ${#instance_devices[@]} devices ($total_bytes bytes raw) for aggregate local I/O"
+elif [[ "$layout" == "single" ]]; then
+  echo "Plan: format ${selected_devices[0]} as XFS and mount it at $mount_point"
+else
+  echo "Plan: create RAID0 across ${#selected_devices[@]} devices ($selected_raw_bytes bytes raw), format as XFS, and mount at $mount_point"
+fi
+
+if (( hazards )); then
+  echo "Refusing to format instance-store devices because the checks above found existing state" >&2
+  exit 1
+fi
+
+if (( layout_decision_required )) && [[ "$mode" == "--apply" ]]; then
+  echo "Refusing to apply an implicit layout; rerun with --layout single or --layout raid0" >&2
+  exit 1
+fi
+
+if [[ "$mode" == "--check" ]]; then
+  echo "Check complete; no storage changes were made"
+  exit 0
+fi
+
+if command -v dnf >/dev/null 2>&1; then
+  package_manager=(sudo dnf install -y)
+elif command -v yum >/dev/null 2>&1; then
+  package_manager=(sudo yum install -y)
+elif command -v apt-get >/dev/null 2>&1; then
+  sudo apt-get update
+  package_manager=(sudo apt-get install -y)
+else
+  echo "No supported package manager found" >&2
+  exit 1
+fi
+
+"${package_manager[@]}" xfsprogs
+sudo mkdir -p "$mount_point"
+
+if [[ "$layout" == "single" ]]; then
+  storage_device="${selected_devices[0]}"
+else
+  "${package_manager[@]}" mdadm
+  storage_device=/dev/md0
+  if [[ -e "$storage_device" ]]; then
+    echo "$storage_device already exists; refusing to replace it" >&2
+    exit 1
+  fi
+  sudo mdadm \
+    --create "$storage_device" \
+    --level=0 \
+    --raid-devices="${#selected_devices[@]}" \
+    --metadata=1.2 \
+    --run \
+    "${selected_devices[@]}"
+fi
+
+sudo mkfs.xfs "$storage_device"
+sudo mount "$storage_device" "$mount_point"
+sudo chown "$(id -u):$(id -g)" "$mount_point"
+mkdir -p "$mount_point/dolma3p5-resharding"
+{
+  printf 'storage_device=%s\n' "$storage_device"
+  printf 'instance_devices=%s\n' "${selected_devices[*]}"
+} > "$marker_path"
+
+available_bytes=$(df -B1 --output=avail "$mount_point" | tail -n 1 | tr -d ' ')
+if (( available_bytes < minimum_available_bytes )); then
+  echo "Prepared storage has only $available_bytes available bytes; minimum is $minimum_available_bytes" >&2
+  exit 1
+fi
+
+echo "Prepared $mount_point with $available_bytes available bytes"
+df -h "$mount_point"
diff --git a/scripts/dolma3p5_resharding/validate.py b/scripts/dolma3p5_resharding/validate.py
new file mode 100644
index 00000000..a8469170
--- /dev/null
+++ b/scripts/dolma3p5_resharding/validate.py
@@ -0,0 +1,34 @@
+"""Validate the locally prepared Dolma 3.5 resharding artifacts."""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+
+from workflow import DEFAULT_BUILD_PATH, PreparationError, validate_build
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description=__doc__,
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
+    )
+    parser.add_argument(
+        "--build",
+        type=Path,
+        default=DEFAULT_BUILD_PATH,
+        help="local preparation build directory",
+    )
+    return parser
+
+
+def main() -> None:
+    parser = build_parser()
+    try:
+        validate_build(parser.parse_args())
+    except PreparationError as exc:
+        parser.exit(2, f"error: {exc}\n")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/dolma3p5_resharding/verify.py b/scripts/dolma3p5_resharding/verify.py
new file mode 100644
index 00000000..1cc36488
--- /dev/null
+++ b/scripts/dolma3p5_resharding/verify.py
@@ -0,0 +1,55 @@
+"""Verify materialized Dolma 3.5 outputs using object sizes only."""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+
+from workflow import DEFAULT_BUILD_PATH, DEFAULT_REGION, PreparationError, verify_output
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description=__doc__,
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
+    )
+    parser.add_argument(
+        "--build",
+        type=Path,
+        default=DEFAULT_BUILD_PATH,
+        help="local preparation build directory",
+    )
+    parser.add_argument("--profile", help="optional AWS profile")
+    parser.add_argument(
+        "--region",
+        default=DEFAULT_REGION,
+        help="AWS region; override when the destination store is in another region",
+    )
+    parser.add_argument("--max-workers", type=int)
+    selection = parser.add_mutually_exclusive_group()
+    selection.add_argument(
+        "--all",
+        action="store_true",
+        help="verify every execution unit; this is also the default selection",
+    )
+    selection.add_argument(
+        "--category",
+        help="verify one exact mix name, leaf ID, or MIX_NAME::CATEGORY_NAME selector",
+    )
+    selection.add_argument(
+        "--unit",
+        help="verify one exact execution-unit ID",
+    )
+    return parser
+
+
+def main() -> None:
+    parser = build_parser()
+    try:
+        verify_output(parser.parse_args())
+    except PreparationError as exc:
+        parser.exit(2, f"error: {exc}\n")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/scripts/dolma3p5_resharding/workflow.py b/scripts/dolma3p5_resharding/workflow.py
new file mode 100644
index 00000000..9dbed6f8
--- /dev/null
+++ b/scripts/dolma3p5_resharding/workflow.py
@@ -0,0 +1,5997 @@
+"""Shared implementation for the Dolma 3.5 resharding preparation scripts.
+
+This module has no command-line interface; each workflow phase has a sibling
+script. It does not materialize token data.
+"""
+
+from __future__ import annotations
+
+import argparse
+import base64
+import copy
+import csv
+import hashlib
+import html
+import json
+import math
+import os
+import posixpath
+import re
+import shlex
+import shutil
+import subprocess
+import time
+from collections import Counter, defaultdict
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from dataclasses import dataclass
+from dataclasses import field as dataclass_field
+from datetime import datetime, timezone
+from functools import lru_cache
+from pathlib import Path
+from typing import Any, Callable, Iterable, Iterator, Sequence
+from urllib.parse import unquote, urlparse
+
+import boto3
+import yaml
+
+try:
+    from .output_report import render_output_validation_report, report_root_style
+except ImportError:
+    from output_report import render_output_validation_report, report_root_style
+
+UINT32_BYTES = 4
+# Mirrors DOCUMENT_SELECTION_ALGORITHM in
+# python/dolma/tokenizer/document_selection.py. Copied rather than imported to keep
+# this module importable with only boto3 and PyYAML installed.
+DOCUMENT_SELECTION_ALGORITHM = "document_hash_bucket_v1"
+# Pins RESHARDING_MANIFEST_SCHEMA_VERSION from python/dolma/tokenizer/reshard.py.
+# The generated launcher makes the worker runtime assert the same value.
+EXPECTED_MANIFEST_SCHEMA_VERSION = 2
+EXECUTION_UNIT_INDEX_WIDTH = 8
+EXECUTION_LAYOUT_SCHEMA_VERSION = 1
+DESTINATION_LAYOUT = "build-scoped-category-output-v1"
+BUILD_ID_PREFIX = "dolma3p5-14t-"
+DEFAULT_TARGET = 14_000_000_000_000
+DEFAULT_REGION = "us-east-1"
+REPOSITORY_ROOT = Path(__file__).resolve().parents[2]
+DEFAULT_BUILD_PATH = REPOSITORY_ROOT / "runs/dolma3p5-resharding/14t"
+PREPARATION_PHASES = (
+    "01-plan",
+    "02-preflight",
+    "03-output-validation",
+)
+# Phases generated after the plan. Replacing any plan stage invalidates all of
+# them, so they are reset together.
+DOWNSTREAM_PREPARATION_PHASES = PREPARATION_PHASES[1:]
+PLAN_STAGES = ("resolution", "inventory", "execution")
+PLAN_ROOT_ARTIFACTS = {"report.html"}
+LEGACY_PREPARATION_PHASES = (
+    "02-inventory",
+    "03-proposal",
+    "04-preflight",
+    "05-output-validation",
+)
+PRESERVED_BUILD_METADATA = {".DS_Store"}
+DEFAULT_SETTINGS: dict[str, Any] = {
+    "target_uint32_values": DEFAULT_TARGET,
+    "direct_s3_bucket": "ai2-llm",
+    "s5cmd_numworkers": 64,
+    "s5cmd_retry_count": 10,
+    "inventory_max_workers": 32,
+    "max_listing_catalog_objects": 5_000,
+    "max_listing_overfetch_ratio": 8.0,
+    "minimum_catalog_prefix_components": 4,
+    "random_seed": 42,
+    "max_workers_per_reshard": 32,
+    "s5cmd_download_concurrency": 32,
+    "tokenizer_name_or_path": "allenai/dolma2-tokenizer",
+    "target_output_shard_bytes": 64 * 1024**3,
+    "max_output_shards_per_unit": 8,
+    "worker_disk_headroom_ratio": 1.1,
+    "document_selection_work_passes": 2,
+    "worker_instance_grid": [
+        {
+            "instance_type": "i4i.2xlarge",
+            "vcpus": 8,
+            "local_nvme_devices": 1,
+            "local_nvme_bytes": 1_875_000_000_000,
+            "max_estimated_work_uint32_values": 150_000_000_000,
+        },
+        {
+            "instance_type": "i4i.4xlarge",
+            "vcpus": 16,
+            "local_nvme_devices": 1,
+            "local_nvme_bytes": 3_750_000_000_000,
+            "max_estimated_work_uint32_values": 300_000_000_000,
+        },
+        {
+            "instance_type": "i4i.8xlarge",
+            "vcpus": 32,
+            "local_nvme_devices": 2,
+            "local_nvme_bytes": 7_500_000_000_000,
+            "max_estimated_work_uint32_values": None,
+        },
+    ],
+    "maximum_expected_upsample_rate": None,
+    "max_materialized_unit_target_residual_fraction": 0.001,
+    "max_materialized_total_target_residual_fraction": 0.00001,
+}
+
+
+class PreparationError(RuntimeError):
+    """A user-actionable preparation or validation failure."""
+
+
+def normalize_region(region: str | None) -> str:
+    """Return the region, or DEFAULT_REGION when it is empty or None."""
+
+    normalized = (region or "").strip()
+    return normalized or DEFAULT_REGION
+
+
+@dataclass(frozen=True)
+class S3Object:
+    bucket: str
+    key: str
+    size_bytes: int
+    etag: str = ""
+    last_modified: str = ""
+    storage_class: str = ""
+    source: str = "listing"
+
+    @property
+    def uri(self) -> str:
+        return f"s3://{self.bucket}/{self.key}"
+
+
+@dataclass(frozen=True)
+class S5cmdRunResult:
+    returncode: int
+    stderr: str
+    output_records: int
+    elapsed_seconds: float
+
+
+def _utc_now() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+def _normalize_timestamp(value: str) -> str:
+    if not value:
+        return ""
+    try:
+        parsed = datetime.fromisoformat(value.replace("Z", "+00:00"))
+    except ValueError:
+        return value
+    if parsed.tzinfo is None:
+        parsed = parsed.replace(tzinfo=timezone.utc)
+    return parsed.astimezone(timezone.utc).isoformat()
+
+
+def _sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as f:
+        for block in iter(lambda: f.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def _count_label(value: int, singular: str, plural: str | None = None) -> str:
+    label = singular if value == 1 else plural or f"{singular}s"
+    return f"{value:,} {label}"
+
+
+def _expected_build_id(mix_sha256: str, catalog_sha256: str) -> str:
+    """Return the build ID for a mix and catalog SHA-256 pair."""
+
+    seed = f"{mix_sha256}:{catalog_sha256}".encode()
+    return f"{BUILD_ID_PREFIX}{hashlib.sha256(seed).hexdigest()[:12]}"
+
+
+def _validate_preparation_build(path: Path) -> dict[str, Any]:
+    """Return build.json after checking the directory belongs to this workflow.
+
+    Raises PreparationError unless build.json is a regular file whose build_id
+    matches its recorded mix and catalog digests, and every top-level entry is a
+    recognized phase.
+    """
+
+    if not path.exists():
+        raise PreparationError(
+            f"No preparation build found at {path}. Run "
+            "scripts/dolma3p5_resharding/plan.py first, or pass the build "
+            "directory of an existing plan."
+        )
+    if path.is_symlink() or not path.is_dir():
+        raise PreparationError(f"Preparation build is not a real directory: {path}")
+    manifest_path = path / "build.json"
+    if manifest_path.is_symlink() or not manifest_path.is_file():
+        raise PreparationError(
+            "Refusing to replace an unrecognized directory. Expected this "
+            f"workflow's build.json marker in: {path}"
+        )
+    try:
+        with manifest_path.open(encoding="utf-8") as f:
+            manifest = json.load(f)
+    except (OSError, json.JSONDecodeError) as exc:
+        raise PreparationError(
+            f"Invalid preparation build marker: {manifest_path}"
+        ) from exc
+    build_id = manifest.get("build_id")
+    mix_sha256 = manifest.get("mix_sha256")
+    catalog_sha256 = manifest.get("catalog_sha256")
+    hashes_are_valid = all(
+        isinstance(value, str) and re.fullmatch(r"[0-9a-f]{64}", value)
+        for value in (mix_sha256, catalog_sha256)
+    )
+    expected_build_id = ""
+    if hashes_are_valid:
+        expected_build_id = _expected_build_id(str(mix_sha256), str(catalog_sha256))
+    if (
+        manifest.get("schema_version") != 1
+        or not isinstance(build_id, str)
+        or build_id != expected_build_id
+    ):
+        raise PreparationError(
+            f"Refusing to replace an unrecognized preparation build: {path}"
+        )
+    recognized_phases = {*PREPARATION_PHASES, *LEGACY_PREPARATION_PHASES}
+    unknown = sorted(
+        child.name
+        for child in path.iterdir()
+        if child.name
+        not in {"build.json", *recognized_phases, *PRESERVED_BUILD_METADATA}
+    )
+    if unknown:
+        raise PreparationError(
+            "Refusing to reset a preparation build containing unknown top-level "
+            f"entries: {', '.join(unknown)}"
+        )
+    for phase_name in recognized_phases:
+        phase = path / phase_name
+        if phase.exists() and (phase.is_symlink() or not phase.is_dir()):
+            raise PreparationError(
+                f"Refusing to replace an unsafe preparation phase path: {phase}"
+            )
+    return manifest
+
+
+def _validate_execution_layout(build: Path) -> dict[str, Any]:
+    """Return 01-plan/execution/dataset-layout.json, raising if it is stale.
+
+    The layout must name this build's ID and dataset root, DESTINATION_LAYOUT,
+    and EXECUTION_UNIT_INDEX_WIDTH.
+    """
+
+    manifest = _validate_preparation_build(build)
+    layout_path = build / "01-plan/execution/dataset-layout.json"
+    if layout_path.is_symlink() or not layout_path.is_file():
+        raise PreparationError(
+            "Execution layout is missing. Rerun scripts/dolma3p5_resharding/plan.py before preflight "
+            "or materialization."
+        )
+    try:
+        layout = json.loads(layout_path.read_text(encoding="utf-8"))
+    except (OSError, json.JSONDecodeError) as exc:
+        raise PreparationError(f"Invalid execution layout: {layout_path}") from exc
+    expected = {
+        "build_id": manifest["build_id"],
+        "layout": DESTINATION_LAYOUT,
+        "execution_unit_index_width": EXECUTION_UNIT_INDEX_WIDTH,
+        "execution_unit_id_width": EXECUTION_UNIT_INDEX_WIDTH,
+    }
+    mismatches = [name for name, value in expected.items() if layout.get(name) != value]
+    dataset_root = layout.get("dataset_root")
+    if not isinstance(dataset_root, str) or not dataset_root.endswith(
+        f"/{manifest['build_id']}"
+    ):
+        mismatches.append("dataset_root")
+    if mismatches:
+        found_layout = layout.get("layout", "missing")
+        raise PreparationError(
+            "Execution plan uses an obsolete destination layout "
+            f"(layout={found_layout}). Rerun "
+            "scripts/dolma3p5_resharding/plan.py before preflight or materialization."
+        )
+    return layout
+
+
+def _remove_generated_phase(path: Path) -> None:
+    if not path.exists():
+        return
+    if path.is_symlink() or not path.is_dir():
+        raise PreparationError(
+            f"Refusing to replace an unsafe preparation phase path: {path}"
+        )
+    shutil.rmtree(path)
+
+
+def _reset_preparation_build(path: Path) -> None:
+    """Remove the generated phases and build.json from a recognized build.
+
+    Creates the directory when it does not exist and leaves an empty one alone.
+    """
+
+    if not path.exists():
+        path.mkdir(parents=True, exist_ok=False)
+        return
+    if path.is_symlink() or not path.is_dir():
+        raise PreparationError(f"Preparation output is not a real directory: {path}")
+    if not any(path.iterdir()):
+        return
+    _validate_preparation_build(path)
+    for phase_name in dict.fromkeys((*PREPARATION_PHASES, *LEGACY_PREPARATION_PHASES)):
+        _remove_generated_phase(path / phase_name)
+    (path / "build.json").unlink()
+
+
+def _reset_preparation_phase(
+    build: Path, phase_name: str, *downstream_phase_names: str
+) -> Path:
+    """Remove phase_name and the downstream phases, then recreate phase_name.
+
+    Every name must be in PREPARATION_PHASES. Returns the new phase directory.
+    """
+
+    _validate_preparation_build(build)
+    names = (phase_name, *downstream_phase_names)
+    if any(name not in PREPARATION_PHASES for name in names):
+        raise ValueError(f"Unknown preparation phase: {names}")
+    for name in names:
+        _remove_generated_phase(build / name)
+    phase = build / phase_name
+    phase.mkdir(exist_ok=False)
+    return phase
+
+
+def _reset_plan_stage(build: Path, stage_name: str, *downstream_stage_names: str) -> Path:
+    """Remove stage_name and the downstream stages, then recreate stage_name.
+
+    Every name must be in PLAN_STAGES. Also removes PLAN_ROOT_ARTIFACTS and
+    every DOWNSTREAM_PREPARATION_PHASES directory. Returns the new stage
+    directory.
+    """
+
+    _validate_preparation_build(build)
+    plan_root = build / "01-plan"
+    if plan_root.is_symlink() or not plan_root.is_dir():
+        raise PreparationError(f"Plan output is not a real directory: {plan_root}")
+    unknown = sorted(
+        child.name
+        for child in plan_root.iterdir()
+        if child.name not in {*PLAN_STAGES, *PLAN_ROOT_ARTIFACTS, *PRESERVED_BUILD_METADATA}
+    )
+    if unknown:
+        raise PreparationError("Refusing to reset a plan containing unknown entries: " + ", ".join(unknown))
+    if stage_name == "execution":
+        _restore_inventory_report_from_combined(plan_root)
+    names = (stage_name, *downstream_stage_names)
+    if any(name not in PLAN_STAGES for name in names):
+        raise ValueError(f"Unknown plan stage: {names}")
+    for artifact_name in PLAN_ROOT_ARTIFACTS:
+        artifact = plan_root / artifact_name
+        if artifact.exists():
+            if artifact.is_symlink() or not artifact.is_file():
+                raise PreparationError(f"Refusing to replace an unsafe plan artifact: {artifact}")
+            artifact.unlink()
+    for name in names:
+        _remove_generated_phase(plan_root / name)
+    for phase_name in DOWNSTREAM_PREPARATION_PHASES:
+        _remove_generated_phase(build / phase_name)
+    stage = plan_root / stage_name
+    stage.mkdir(exist_ok=False)
+    return stage
+
+
+def _restore_inventory_report_from_combined(plan_root: Path) -> None:
+    """Restore 01-plan/inventory/report.html from the combined 01-plan/report.html.
+
+    Does nothing when the inventory report is already present or the combined
+    report is absent.
+    """
+
+    inventory_report = plan_root / "inventory/report.html"
+    if inventory_report.is_file() and not inventory_report.is_symlink():
+        return
+    combined_report = plan_root / "report.html"
+    if combined_report.is_symlink() or not combined_report.is_file():
+        return
+    combined = combined_report.read_text(encoding="utf-8")
+    match = re.search(
+        r'<template id="source-report-document">(?P<document>.*?)</template>\s*'
+        r'<template id="execution-report-document">',
+        combined,
+        flags=re.DOTALL,
+    )
+    if match is None:
+        raise PreparationError(f"Cannot recover the inventory report from: {combined_report}")
+    document = match.group("document").replace('<head><base href="inventory/">', "<head>", 1)
+    _write_text(inventory_report, document)
+
+
+def _write_text(path: Path, value: str) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    try:
+        with path.open("x", encoding="utf-8") as f:
+            f.write(value)
+    except FileExistsError as exc:
+        raise PreparationError(
+            f"Refusing to replace existing artifact: {path}"
+        ) from exc
+
+
+def _write_json(path: Path, value: Any) -> None:
+    _write_text(path, json.dumps(value, indent=2, sort_keys=True) + "\n")
+
+
+def _write_csv(
+    path: Path, rows: Iterable[dict[str, Any]], fieldnames: Sequence[str]
+) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    try:
+        f = path.open("x", newline="", encoding="utf-8")
+    except FileExistsError as exc:
+        raise PreparationError(
+            f"Refusing to replace existing artifact: {path}"
+        ) from exc
+    with f:
+        writer = csv.DictWriter(f, fieldnames=fieldnames, extrasaction="ignore")
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def _read_csv(path: Path) -> list[dict[str, str]]:
+    with path.open(newline="", encoding="utf-8") as f:
+        return list(csv.DictReader(f))
+
+
+def _load_settings(path: Path | None) -> dict[str, Any]:
+    settings = copy.deepcopy(DEFAULT_SETTINGS)
+    if path is not None:
+        with path.open(encoding="utf-8") as f:
+            loaded = yaml.safe_load(f) or {}
+        if not isinstance(loaded, dict):
+            raise PreparationError(f"Settings must be a YAML mapping: {path}")
+        unknown = sorted(set(loaded) - set(DEFAULT_SETTINGS))
+        if unknown:
+            raise PreparationError(f"Unknown settings: {', '.join(unknown)}")
+        settings.update(loaded)
+    for name in (
+        "target_uint32_values",
+        "max_workers_per_reshard",
+        "s5cmd_download_concurrency",
+        "target_output_shard_bytes",
+        "max_output_shards_per_unit",
+        "inventory_max_workers",
+        "s5cmd_numworkers",
+        "minimum_catalog_prefix_components",
+        "max_listing_catalog_objects",
+    ):
+        if int(settings[name]) <= 0:
+            raise PreparationError(f"{name} must be positive")
+    if float(settings["max_listing_overfetch_ratio"]) <= 0:
+        raise PreparationError("max_listing_overfetch_ratio must be positive")
+    if float(settings["worker_disk_headroom_ratio"]) <= 1:
+        raise PreparationError("worker_disk_headroom_ratio must be greater than one")
+    for name in ("document_selection_work_passes", "s5cmd_retry_count"):
+        if int(settings[name]) < 0:
+            raise PreparationError(f"{name} cannot be negative")
+    settings["worker_instance_grid"] = _validate_worker_instance_grid(settings["worker_instance_grid"])
+    for name in (
+        "max_materialized_unit_target_residual_fraction",
+        "max_materialized_total_target_residual_fraction",
+    ):
+        value = float(settings[name])
+        if not 0 < value < 1:
+            raise PreparationError(f"{name} must be between zero and one")
+    maximum_expected_upsample_rate = settings["maximum_expected_upsample_rate"]
+    if maximum_expected_upsample_rate is not None and float(maximum_expected_upsample_rate) <= 1:
+        raise PreparationError("maximum_expected_upsample_rate must be greater than one")
+    return settings
+
+
+def _validate_worker_instance_grid(value: Any) -> list[dict[str, Any]]:
+    """Return the normalized worker_instance_grid.
+
+    Entries must be distinct i4i types ordered by strictly increasing vCPUs,
+    local NVMe bytes, and work limits. Only the final entry may set
+    max_estimated_work_uint32_values to None, which means unbounded work.
+    """
+
+    if not isinstance(value, list) or not value:
+        raise PreparationError("worker_instance_grid must be a non-empty list")
+    required = {
+        "instance_type",
+        "vcpus",
+        "local_nvme_devices",
+        "local_nvme_bytes",
+        "max_estimated_work_uint32_values",
+    }
+    grid: list[dict[str, Any]] = []
+    seen: set[str] = set()
+    previous_vcpus = 0
+    previous_storage = 0
+    previous_work_limit = 0
+    for index, raw in enumerate(value):
+        if not isinstance(raw, dict):
+            raise PreparationError(f"worker_instance_grid entry {index + 1} must be a mapping")
+        missing = required - set(raw)
+        unknown = set(raw) - required
+        if missing or unknown:
+            details = []
+            if missing:
+                details.append("missing " + ", ".join(sorted(missing)))
+            if unknown:
+                details.append("unknown " + ", ".join(sorted(unknown)))
+            raise PreparationError(f"Invalid worker_instance_grid entry {index + 1}: " + "; ".join(details))
+        instance_type = str(raw["instance_type"]).strip()
+        if not re.fullmatch(r"i4i\.[A-Za-z0-9]+", instance_type):
+            raise PreparationError(
+                "worker_instance_grid is restricted to i4i instance types; "
+                f"found {instance_type or 'empty value'}"
+            )
+        if instance_type in seen:
+            raise PreparationError(f"Duplicate worker instance type in grid: {instance_type}")
+        seen.add(instance_type)
+        vcpus = int(raw["vcpus"])
+        devices = int(raw["local_nvme_devices"])
+        storage = int(raw["local_nvme_bytes"])
+        raw_work_limit = raw["max_estimated_work_uint32_values"]
+        work_limit = int(raw_work_limit) if raw_work_limit is not None else None
+        if vcpus <= 0 or devices <= 0 or storage <= 0:
+            raise PreparationError(f"Worker resources must be positive for {instance_type}")
+        if work_limit is not None and work_limit <= 0:
+            raise PreparationError(f"Work limit must be positive for {instance_type}")
+        if vcpus <= previous_vcpus or storage <= previous_storage:
+            raise PreparationError("worker_instance_grid must increase in both vCPUs and local NVMe")
+        if work_limit is None:
+            if index != len(value) - 1:
+                raise PreparationError("Only the final worker_instance_grid entry may have no work limit")
+        elif work_limit <= previous_work_limit:
+            raise PreparationError("worker_instance_grid work limits must increase with instance size")
+        grid.append(
+            {
+                "instance_type": instance_type,
+                "vcpus": vcpus,
+                "local_nvme_devices": devices,
+                "local_nvme_bytes": storage,
+                "max_estimated_work_uint32_values": work_limit,
+            }
+        )
+        previous_vcpus = vcpus
+        previous_storage = storage
+        if work_limit is not None:
+            previous_work_limit = work_limit
+    if grid[-1]["max_estimated_work_uint32_values"] is not None:
+        raise PreparationError("The final worker_instance_grid entry must accept unbounded work")
+    return grid
+
+
+def _load_catalog(path: Path) -> list[dict[str, str]]:
+    rows: list[dict[str, str]] = []
+    with path.open(newline="", encoding="utf-8-sig") as f:
+        reader = csv.reader(f)
+        for line_number, row in enumerate(reader, start=1):
+            if not row or all(not value.strip() for value in row):
+                continue
+            if len(row) < 2:
+                raise PreparationError(
+                    f"Catalog row {line_number} has fewer than two columns"
+                )
+            bucket, encoded_key = row[0].strip(), row[1].strip()
+            if (
+                line_number == 1
+                and bucket.lower() == "bucket"
+                and encoded_key.lower() in {"key", "path"}
+            ):
+                continue
+            key = unquote(encoded_key)
+            if not bucket:
+                raise PreparationError(
+                    f"Invalid catalog object on row {line_number}: {row[:2]}"
+                )
+            if any(ord(char) < 32 for char in bucket + key):
+                raise PreparationError(
+                    f"Control character in catalog object on row {line_number}"
+                )
+            if not key.endswith(".npy"):
+                # The catalog also carries metadata rows. Only NPY rows define
+                # membership; the derived metadata partner is verified against
+                # S3 during inventory.
+                continue
+            rows.append(
+                {"bucket": bucket, "key": key, "catalog_line": str(line_number)}
+            )
+    if not rows:
+        raise PreparationError(f"Catalog contains no NPY objects: {path}")
+    return rows
+
+
+def _catalog_pattern(yaml_path: str) -> str:
+    relative = yaml_path.removeprefix("dolma3p5_pool/")
+    return (
+        relative if relative.startswith("preprocessed/") else f"preprocessed/{relative}"
+    )
+
+
+def _direct_s3_pattern(yaml_path: str, bucket: str) -> tuple[str, str]:
+    if not yaml_path.startswith("preprocessed/"):
+        raise PreparationError(f"Unsupported direct path: {yaml_path}")
+    return bucket, yaml_path
+
+
+@lru_cache(maxsize=None)
+def _compiled_path_pattern(pattern: str) -> re.Pattern[str]:
+    # "*" is the only wildcard in a YAML path. Every other character is escaped,
+    # so names containing regex or glob metacharacters match literally.
+    expression = re.escape(unquote(pattern)).replace(r"\*", ".*")
+    return re.compile(expression)
+
+
+def _matches_key(key: str, pattern: str) -> bool:
+    return _compiled_path_pattern(pattern).fullmatch(unquote(key)) is not None
+
+
+def _literal_prefix(pattern: str) -> str:
+    wildcard_offset = pattern.find("*")
+    prefix = pattern[:wildcard_offset] if wildcard_offset >= 0 else pattern
+    if not prefix.endswith("/"):
+        prefix = prefix.rsplit("/", 1)[0] + "/"
+    return prefix
+
+
+def _common_directory_prefix(keys: Sequence[str]) -> str:
+    if not keys:
+        raise ValueError("Cannot find a common prefix for no keys")
+    directories = [key.rsplit("/", 1)[0] for key in keys]
+    common = posixpath.commonpath(directories)
+    return common.rstrip("/") + "/"
+
+
+def _pair_metadata_key(npy_key: str) -> str:
+    if not npy_key.endswith(".npy"):
+        raise ValueError(npy_key)
+    return npy_key[:-4] + ".csv.gz"
+
+
+def _build_id(mix_path: Path, catalog_path: Path) -> str:
+    return _expected_build_id(_sha256(mix_path), _sha256(catalog_path))
+
+
+@dataclass
+class _MixResolution:
+    """Resolution-stage output for one mix YAML document."""
+
+    normalized_mix: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    normalized_paths: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    catalog_matches: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    direct_patterns: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    corrections: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    failures: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    duplicates: list[dict[str, Any]] = dataclass_field(default_factory=list)
+    effective_weight_total: float = 0.0
+
+
+def _resolve_category_paths(
+    resolution: _MixResolution,
+    category: dict[str, Any],
+    *,
+    leaf_id: str,
+    mix_index: int,
+    mix_name: str,
+    category_index: int,
+    category_name: str,
+    active: bool,
+    catalog: Sequence[dict[str, str]],
+    direct_s3_bucket: str,
+) -> None:
+    """Resolve one category's YAML paths into the fields of resolution.
+
+    A dolma3p5_pool/ path is matched against the catalog, a preprocessed/ path
+    becomes a direct S3 pattern, and any other root becomes a failure row.
+    """
+
+    seen_paths: Counter[str] = Counter()
+    for path_index, yaml_path_value in enumerate(category.get("paths", [])):
+        yaml_path = str(yaml_path_value)
+        if any(ord(char) < 32 for char in yaml_path):
+            resolution.failures.append(
+                {
+                    "mix_name": mix_name,
+                    "category_name": category_name,
+                    "path": repr(yaml_path),
+                    "reason": "control character in YAML path",
+                }
+            )
+            continue
+        seen_paths[yaml_path] += 1
+        drop_tcl_duplicate = (
+            mix_name == "the-stack-v2:Tcl"
+            and category_name == "high"
+            and "/quality_p95/" in yaml_path
+            and seen_paths[yaml_path] > 1
+        )
+        if seen_paths[yaml_path] > 1:
+            resolution.duplicates.append(
+                {
+                    "leaf_id": leaf_id,
+                    "mix_name": mix_name,
+                    "category_name": category_name,
+                    "path": yaml_path,
+                    "occurrence": seen_paths[yaml_path],
+                    "action": "dropped" if drop_tcl_duplicate else "retained",
+                }
+            )
+        if drop_tcl_duplicate:
+            resolution.corrections.append(
+                {
+                    "type": "drop_duplicate",
+                    "mix_name": mix_name,
+                    "category_name": category_name,
+                    "path": yaml_path,
+                    "reason": "explicitly approved duplicate Tcl quality_p95 correction",
+                }
+            )
+            continue
+
+        path_id = f"{leaf_id}:{path_index:03d}"
+        resolution_route = (
+            "catalog"
+            if yaml_path.startswith("dolma3p5_pool/")
+            else "direct_s3"
+            if yaml_path.startswith("preprocessed/")
+            else "unsupported"
+        )
+        resolution.normalized_paths.append(
+            {
+                "path_id": path_id,
+                "leaf_id": leaf_id,
+                "mix_index": mix_index,
+                "mix_name": mix_name,
+                "category_index": category_index,
+                "category_name": category_name,
+                "active": str(active).lower(),
+                "yaml_path": yaml_path,
+            }
+        )
+        if resolution_route == "catalog":
+            pattern = _catalog_pattern(yaml_path)
+            matched = [row for row in catalog if _matches_key(row["key"], pattern)]
+            if not matched:
+                resolution.failures.append(
+                    {
+                        "mix_name": mix_name,
+                        "category_name": category_name,
+                        "path": yaml_path,
+                        "reason": "no matching object in reference catalog",
+                    }
+                )
+            for match in matched:
+                resolution.catalog_matches.append(
+                    {
+                        "path_id": path_id,
+                        "leaf_id": leaf_id,
+                        "mix_index": mix_index,
+                        "mix_name": mix_name,
+                        "category_index": category_index,
+                        "category_name": category_name,
+                        "active": str(active).lower(),
+                        "yaml_path": yaml_path,
+                        "bucket": match["bucket"],
+                        "key": match["key"],
+                        "catalog_line": match["catalog_line"],
+                    }
+                )
+        elif resolution_route == "direct_s3":
+            bucket, pattern = _direct_s3_pattern(yaml_path, direct_s3_bucket)
+            resolution.direct_patterns.append(
+                {
+                    "path_id": path_id,
+                    "leaf_id": leaf_id,
+                    "mix_index": mix_index,
+                    "mix_name": mix_name,
+                    "category_index": category_index,
+                    "category_name": category_name,
+                    "active": str(active).lower(),
+                    "yaml_path": yaml_path,
+                    "bucket": bucket,
+                    "key_pattern": pattern,
+                    "listing_prefix": _literal_prefix(pattern),
+                }
+            )
+        else:
+            resolution.failures.append(
+                {
+                    "mix_name": mix_name,
+                    "category_name": category_name,
+                    "path": yaml_path,
+                    "reason": "unsupported path root",
+                }
+            )
+
+
+def _resolve_mix_document(
+    document: dict[str, Any],
+    catalog: Sequence[dict[str, str]],
+    settings: dict[str, Any],
+) -> _MixResolution:
+    """Normalize the mix into per-category targets and resolved source objects.
+
+    Problems are collected in the returned failures list rather than raised.
+    """
+
+    resolution = _MixResolution()
+    target_total = int(settings["target_uint32_values"])
+    direct_s3_bucket = str(settings["direct_s3_bucket"])
+
+    for mix_index, mix in enumerate(document["mix"]):
+        mix_name = str(mix["name"])
+        mix_weight = float(mix["weight"])
+        categories = mix.get("categories")
+        if not isinstance(categories, list) or not categories:
+            resolution.failures.append(
+                {
+                    "mix_name": mix_name,
+                    "category_name": "",
+                    "path": "",
+                    "reason": "no categories",
+                }
+            )
+            continue
+        category_weight_sum = sum(float(category["weight"]) for category in categories)
+        if not math.isclose(category_weight_sum, 1.0, rel_tol=0.0, abs_tol=1e-8):
+            resolution.failures.append(
+                {
+                    "mix_name": mix_name,
+                    "category_name": "",
+                    "path": "",
+                    "reason": f"category weights sum to {category_weight_sum}",
+                }
+            )
+
+        for category_index, category in enumerate(categories):
+            category_name = str(category["name"])
+            category_weight = float(category["weight"])
+            effective_weight = mix_weight * category_weight
+            resolution.effective_weight_total += effective_weight
+            active = effective_weight > 0
+            leaf_id = f"{mix_index:03d}:{category_index:02d}"
+            resolution.normalized_mix.append(
+                {
+                    "leaf_id": leaf_id,
+                    "mix_index": mix_index,
+                    "mix_name": mix_name,
+                    "mix_weight": f"{mix_weight:.16g}",
+                    "category_index": category_index,
+                    "category_name": category_name,
+                    "category_weight": f"{category_weight:.16g}",
+                    "effective_weight": f"{effective_weight:.16g}",
+                    "target_uint32_values": round(target_total * effective_weight),
+                    "repetition_factor_raw": category.get("repetition_factor", ""),
+                    "active": str(active).lower(),
+                }
+            )
+            _resolve_category_paths(
+                resolution,
+                category,
+                leaf_id=leaf_id,
+                mix_index=mix_index,
+                mix_name=mix_name,
+                category_index=category_index,
+                category_name=category_name,
+                active=active,
+                catalog=catalog,
+                direct_s3_bucket=direct_s3_bucket,
+            )
+
+    mix_weight_sum = sum(float(mix["weight"]) for mix in document["mix"])
+    if not math.isclose(mix_weight_sum, 1.0, rel_tol=0.0, abs_tol=1e-8):
+        resolution.failures.append(
+            {
+                "mix_name": "",
+                "category_name": "",
+                "path": "",
+                "reason": f"mix weights sum to {mix_weight_sum}",
+            }
+        )
+    if not math.isclose(resolution.effective_weight_total, 1.0, rel_tol=0.0, abs_tol=1e-8):
+        resolution.failures.append(
+            {
+                "mix_name": "",
+                "category_name": "",
+                "path": "",
+                "reason": f"effective weights sum to {resolution.effective_weight_total}",
+            }
+        )
+    return resolution
+
+
+def _plan_bulk_listings(
+    catalog_matches: Sequence[dict[str, Any]],
+    direct_patterns: Sequence[dict[str, Any]],
+    catalog_by_bucket: dict[str, list[dict[str, str]]],
+    settings: dict[str, Any],
+) -> list[dict[str, Any]]:
+    """Group resolved objects into S3 listing prefixes.
+
+    A group is split when its common prefix has fewer than
+    minimum_catalog_prefix_components components, or when it would exceed
+    max_listing_catalog_objects or max_listing_overfetch_ratio.
+    """
+
+    listing_groups: dict[tuple[str, str], dict[str, Any]] = {}
+    by_leaf_bucket: dict[tuple[str, str], list[dict[str, Any]]] = defaultdict(list)
+    for match in catalog_matches:
+        by_leaf_bucket[(match["leaf_id"], match["bucket"])].append(match)
+    for (leaf_id, bucket), matches in by_leaf_bucket.items():
+        prefix = _common_directory_prefix([row["key"] for row in matches])
+        prefix_depth = len([part for part in prefix.split("/") if part])
+        if prefix_depth < int(settings["minimum_catalog_prefix_components"]):
+            by_path: dict[str, list[dict[str, Any]]] = defaultdict(list)
+            for match in matches:
+                by_path[match["path_id"]].append(match)
+            subgroups = list(by_path.values())
+        else:
+            subgroups = [matches]
+        for subgroup in subgroups:
+            subgroup_prefix = _common_directory_prefix([row["key"] for row in subgroup])
+            required = {(row["bucket"], row["key"]) for row in subgroup}
+            estimated = sum(
+                1
+                for row in catalog_by_bucket[bucket]
+                if row["key"].startswith(subgroup_prefix)
+            )
+            overfetch = estimated / max(1, len(required))
+            if (
+                estimated > int(settings["max_listing_catalog_objects"])
+                or overfetch > float(settings["max_listing_overfetch_ratio"])
+            ) and len(subgroup) > 1:
+                final_groups = [[row] for row in subgroup]
+            else:
+                final_groups = [subgroup]
+            for final_group in final_groups:
+                final_prefix = _common_directory_prefix(
+                    [row["key"] for row in final_group]
+                )
+                key = (bucket, final_prefix)
+                entry = listing_groups.setdefault(
+                    key,
+                    {
+                        "bucket": bucket,
+                        "listing_prefix": final_prefix,
+                        "leaf_ids": set(),
+                    },
+                )
+                entry["leaf_ids"].update(row["leaf_id"] for row in final_group)
+
+    for row in direct_patterns:
+        key = (row["bucket"], row["listing_prefix"])
+        entry = listing_groups.setdefault(
+            key,
+            {
+                "bucket": row["bucket"],
+                "listing_prefix": row["listing_prefix"],
+                "leaf_ids": set(),
+            },
+        )
+        entry["leaf_ids"].add(row["leaf_id"])
+
+    listing_plan: list[dict[str, Any]] = []
+    for listing_id, ((bucket, prefix), entry) in enumerate(
+        sorted(listing_groups.items())
+    ):
+        estimated = sum(
+            1 for row in catalog_by_bucket[bucket] if row["key"].startswith(prefix)
+        )
+        required = {
+            (row["bucket"], row["key"])
+            for row in catalog_matches
+            if row["bucket"] == bucket and row["key"].startswith(prefix)
+        }
+        listing_plan.append(
+            {
+                "listing_id": listing_id,
+                "bucket": bucket,
+                "listing_prefix": prefix,
+                "leaf_ids": ";".join(sorted(entry["leaf_ids"])),
+                "required_catalog_npy_count": len(required),
+                "estimated_catalog_npy_count": estimated,
+                "estimated_overfetch_ratio": f"{estimated / max(1, len(required)):.6f}",
+            }
+        )
+    return listing_plan
+
+
+def plan_build(args: argparse.Namespace) -> None:
+    """Resolve the mix YAML against the reference catalog into a fresh build.
+
+    Writes build.json and the 01-plan/resolution stage: the normalized mix and
+    path CSVs, catalog-matches.csv, direct-s3-patterns.csv, listing-plan.csv,
+    bulk-listing-commands.txt, the corrections and failure CSVs, and report.html.
+    Raises PreparationError when the resolution produced any failure row.
+    """
+
+    mix_path = args.mix.resolve()
+    catalog_path = args.catalog.resolve()
+    output = args.output.resolve()
+    settings = _load_settings(args.settings.resolve() if args.settings else None)
+
+    with mix_path.open(encoding="utf-8") as f:
+        document = yaml.safe_load(f)
+    if not isinstance(document, dict) or not isinstance(document.get("mix"), list):
+        raise PreparationError("Mix YAML must contain a top-level 'mix' list")
+    catalog = _load_catalog(catalog_path)
+    catalog_by_bucket: dict[str, list[dict[str, str]]] = defaultdict(list)
+    for row in catalog:
+        catalog_by_bucket[row["bucket"]].append(row)
+
+    resolution = _resolve_mix_document(document, catalog, settings)
+    normalized_mix = resolution.normalized_mix
+    normalized_paths = resolution.normalized_paths
+    catalog_matches = resolution.catalog_matches
+    direct_patterns = resolution.direct_patterns
+    corrections = resolution.corrections
+    failures = resolution.failures
+    duplicates = resolution.duplicates
+    listing_plan = _plan_bulk_listings(
+        catalog_matches, direct_patterns, catalog_by_bucket, settings
+    )
+
+    build_id = _build_id(mix_path, catalog_path)
+    manifest = {
+        "schema_version": 1,
+        "build_id": build_id,
+        "created_at": _utc_now(),
+        "mix_path": str(mix_path),
+        "mix_sha256": _sha256(mix_path),
+        "catalog_path": str(catalog_path),
+        "catalog_sha256": _sha256(catalog_path),
+        "settings": settings,
+        "counts": {
+            "mix_entries": len(document["mix"]),
+            "category_entries": len(normalized_mix),
+            "active_categories": sum(row["active"] == "true" for row in normalized_mix),
+            "normalized_paths": len(normalized_paths),
+            "catalog_matches": len(catalog_matches),
+            "direct_patterns": len(direct_patterns),
+            "listing_commands": len(listing_plan),
+            "corrections": len(corrections),
+            "failures": len(failures),
+        },
+    }
+    _reset_preparation_build(output)
+    plan_root = output / "01-plan"
+    plan_root.mkdir(exist_ok=False)
+    phase = plan_root / "resolution"
+    phase.mkdir(exist_ok=False)
+    _write_json(output / "build.json", manifest)
+    _write_csv(
+        phase / "normalized-mix.csv",
+        normalized_mix,
+        list(normalized_mix[0]) if normalized_mix else [],
+    )
+    _write_csv(
+        phase / "normalized-paths.csv",
+        normalized_paths,
+        list(normalized_paths[0]) if normalized_paths else [],
+    )
+    _write_csv(
+        phase / "catalog-matches.csv",
+        catalog_matches,
+        [
+            "path_id",
+            "leaf_id",
+            "mix_index",
+            "mix_name",
+            "category_index",
+            "category_name",
+            "active",
+            "yaml_path",
+            "bucket",
+            "key",
+            "catalog_line",
+        ],
+    )
+    _write_csv(
+        phase / "direct-s3-patterns.csv",
+        direct_patterns,
+        [
+            "path_id",
+            "leaf_id",
+            "mix_index",
+            "mix_name",
+            "category_index",
+            "category_name",
+            "active",
+            "yaml_path",
+            "bucket",
+            "key_pattern",
+            "listing_prefix",
+        ],
+    )
+    _write_csv(
+        phase / "listing-plan.csv",
+        listing_plan,
+        list(listing_plan[0]) if listing_plan else [],
+    )
+    _write_csv(
+        phase / "corrections.csv",
+        corrections,
+        ["type", "mix_name", "category_name", "path", "reason"],
+    )
+    _write_csv(
+        phase / "duplicate-paths.csv",
+        duplicates,
+        ["leaf_id", "mix_name", "category_name", "path", "occurrence", "action"],
+    )
+    _write_csv(
+        phase / "resolution-failures.csv",
+        failures,
+        ["mix_name", "category_name", "path", "reason"],
+    )
+
+    commands = [
+        "ls --etag --storage-class "
+        + shlex.quote("s3://" + row["bucket"] + "/" + row["listing_prefix"] + "*")
+        for row in listing_plan
+    ]
+    _write_text(
+        phase / "bulk-listing-commands.txt",
+        "\n".join(commands) + ("\n" if commands else ""),
+    )
+    _render_plan_report(
+        phase,
+        normalized_mix=normalized_mix,
+        normalized_paths=normalized_paths,
+        catalog_matches=catalog_matches,
+        direct_patterns=direct_patterns,
+    )
+
+    catalog_summary = _count_label(
+        len(catalog_matches), "catalog NPY match", "catalog NPY matches"
+    )
+    pattern_summary = _count_label(len(direct_patterns), "direct S3 pattern")
+    correction_summary = _count_label(len(corrections), "correction")
+    failure_summary = _count_label(len(failures), "blocking failure")
+    print(
+        f"Plan summary: {catalog_summary}, {pattern_summary}, "
+        f"{correction_summary}, {failure_summary}"
+    )
+    if failures:
+        raise PreparationError(
+            f"Plan contains {len(failures)} validation failure(s); inspect {phase / 'resolution-failures.csv'}"
+        )
+    print(f"Created preparation plan: {output}")
+
+
+def _load_build(build: Path) -> dict[str, Any]:
+    build = build.resolve()
+    manifest_path = build / "build.json"
+    if not manifest_path.is_file():
+        raise PreparationError(
+            f"No preparation build found at {build}. Run "
+            "scripts/dolma3p5_resharding/plan.py first, or pass the build "
+            "directory of an existing plan."
+        )
+    try:
+        with manifest_path.open(encoding="utf-8") as f:
+            manifest = json.load(f)
+    except (OSError, json.JSONDecodeError) as exc:
+        raise PreparationError(f"Invalid preparation build marker: {manifest_path}") from exc
+    mix_path = Path(manifest["mix_path"])
+    catalog_path = Path(manifest["catalog_path"])
+    if _sha256(mix_path) != manifest["mix_sha256"]:
+        raise PreparationError(f"Mix YAML changed since plan creation: {mix_path}")
+    if _sha256(catalog_path) != manifest["catalog_sha256"]:
+        raise PreparationError(f"Catalog changed since plan creation: {catalog_path}")
+    settings = dict(DEFAULT_SETTINGS)
+    settings.update(manifest.get("settings", {}))
+    manifest["settings"] = settings
+    return manifest
+
+
+def _list_prefix(client: Any, bucket: str, prefix: str) -> list[S3Object]:
+    objects: list[S3Object] = []
+    paginator = client.get_paginator("list_objects_v2")
+    for page in paginator.paginate(Bucket=bucket, Prefix=prefix):
+        for item in page.get("Contents", []):
+            objects.append(
+                S3Object(
+                    bucket=bucket,
+                    key=item["Key"],
+                    size_bytes=int(item["Size"]),
+                    etag=str(item.get("ETag", "")).strip('"'),
+                    last_modified=str(item.get("LastModified", "")),
+                    storage_class=str(item.get("StorageClass", "")),
+                    source="listing",
+                )
+            )
+    return objects
+
+
+def _head_object(client: Any, bucket: str, key: str) -> S3Object:
+    response = client.head_object(Bucket=bucket, Key=key)
+    return S3Object(
+        bucket=bucket,
+        key=key,
+        size_bytes=int(response["ContentLength"]),
+        etag=str(response.get("ETag", "")).strip('"'),
+        last_modified=str(response.get("LastModified", "")),
+        storage_class=str(response.get("StorageClass", "")),
+        source="head",
+    )
+
+
+def _s3_reader(
+    args: argparse.Namespace, manifest: dict[str, Any], region: str
+) -> tuple[Any, int]:
+    """Return an S3 client and the worker count for concurrent listings.
+
+    The worker count is args.max_workers, or the inventory_max_workers setting.
+    """
+
+    session = (
+        boto3.Session(profile_name=args.profile) if args.profile else boto3.Session()
+    )
+    client = session.client("s3", region_name=region)
+    max_workers = args.max_workers or int(manifest["settings"]["inventory_max_workers"])
+    return client, max_workers
+
+
+def _head_missing_objects(
+    client: Any,
+    identities: Sequence[tuple[str, str]],
+    found: dict[tuple[str, str], S3Object],
+    errors: list[dict[str, str]],
+    *,
+    max_workers: int,
+    on_progress: Callable[[int, int], None] | None = None,
+) -> None:
+    """HeadObject every identity absent from the bulk listings.
+
+    Resolved objects are added to found. Per-object failures are appended to
+    errors rather than raised.
+    """
+
+    total = len(identities)
+    completed = 0
+    with ThreadPoolExecutor(max_workers=max_workers) as pool:
+        futures = {
+            pool.submit(_head_object, client, bucket, key): (bucket, key)
+            for bucket, key in identities
+        }
+        for future in as_completed(futures):
+            bucket, key = futures[future]
+            try:
+                found[(bucket, key)] = future.result()
+            except Exception as exc:
+                errors.append(
+                    {
+                        "operation": "head",
+                        "bucket": bucket,
+                        "key": key,
+                        "error": repr(exc),
+                    }
+                )
+            completed += 1
+            if on_progress is not None:
+                on_progress(completed, total)
+
+
+def _inventory_status(step: int, message: str) -> None:
+    print(f"[inventory {step}/4] {message}", flush=True)
+
+
+def _count_newlines(path: Path, offset: int) -> tuple[int, int]:
+    count = 0
+    with path.open("rb") as f:
+        f.seek(offset)
+        while chunk := f.read(1024 * 1024):
+            count += chunk.count(b"\n")
+        return f.tell(), count
+
+
+def _run_s5cmd_inventory(
+    command: Sequence[str],
+    raw_output: Path,
+    environment: dict[str, str],
+    status: Callable[[str], None],
+    status_interval_seconds: float = 10.0,
+) -> S5cmdRunResult:
+    stderr_path = raw_output.with_name("collector-stderr.txt")
+    started = time.monotonic()
+    output_records = 0
+    output_offset = 0
+    with raw_output.open("x") as stdout, stderr_path.open("x") as stderr:
+        process = subprocess.Popen(
+            list(command),
+            text=True,
+            stdout=stdout,
+            stderr=stderr,
+            env=environment,
+        )
+        while True:
+            try:
+                returncode = process.wait(timeout=status_interval_seconds)
+                break
+            except subprocess.TimeoutExpired:
+                output_offset, new_records = _count_newlines(raw_output, output_offset)
+                output_records += new_records
+                status(
+                    f"Running: {output_records:,} JSON records received, "
+                    f"{time.monotonic() - started:,.0f}s elapsed"
+                )
+    output_offset, new_records = _count_newlines(raw_output, output_offset)
+    output_records += new_records
+    del output_offset
+    stderr_text = stderr_path.read_text()
+    stderr_path.unlink()
+    return S5cmdRunResult(
+        returncode=returncode,
+        stderr=stderr_text,
+        output_records=output_records,
+        elapsed_seconds=time.monotonic() - started,
+    )
+
+
+def collect_inventory(args: argparse.Namespace) -> None:
+    """List and size every S3 object the resolved plan requires.
+
+    Requires s5cmd on PATH. Writes the 01-plan/inventory stage:
+    raw-listings.jsonl, collector.json, normalized-s3-inventory.csv,
+    required-objects.csv, the missing-object and error CSVs,
+    sampling-rate-audit.csv, inventory-summary.json, inventory-details.json,
+    and report.html.
+    """
+
+    build = args.build.resolve()
+    manifest = _load_build(build)
+    region = normalize_region(args.region)
+    if shutil.which("s5cmd") is None:
+        raise PreparationError(
+            "s5cmd is required for inventory collection and was not found on PATH"
+        )
+    phase = _reset_plan_stage(build, "inventory", "execution")
+    listing_plan = _read_csv(build / "01-plan/resolution/listing-plan.csv")
+
+    client, max_workers = _s3_reader(args, manifest, region)
+    listed: dict[tuple[str, str], S3Object] = {}
+    errors: list[dict[str, str]] = []
+    raw_output = phase / "raw-listings.jsonl"
+    commands_path = build / "01-plan/resolution/bulk-listing-commands.txt"
+    command = [
+        "s5cmd",
+        "--json",
+        "--numworkers",
+        str(manifest["settings"]["s5cmd_numworkers"]),
+        "--retry-count",
+        str(manifest["settings"]["s5cmd_retry_count"]),
+        "run",
+        str(commands_path),
+    ]
+    environment = os.environ.copy()
+    if args.profile:
+        environment["AWS_PROFILE"] = args.profile
+    environment["AWS_REGION"] = region
+    environment["AWS_DEFAULT_REGION"] = region
+    _inventory_status(
+        1,
+        f"Bulk listing: {len(listing_plan):,} commands, "
+        f"{int(manifest['settings']['s5cmd_numworkers']):,} s5cmd workers, "
+        f"{int(manifest['settings']['s5cmd_retry_count']):,} retries",
+    )
+    result = _run_s5cmd_inventory(
+        command,
+        raw_output,
+        environment,
+        lambda message: _inventory_status(1, message),
+    )
+    _write_json(
+        phase / "collector.json",
+        {
+            "collector": "s5cmd",
+            "command": command,
+            "command_count": len(listing_plan),
+            "output_records": result.output_records,
+            "elapsed_seconds": round(result.elapsed_seconds, 3),
+            "raw_output": raw_output.name,
+            "read_only": True,
+        },
+    )
+    listed, parse_errors = _parse_s5cmd_jsonl(raw_output)
+    errors.extend(
+        {
+            "operation": "parse",
+            "bucket": "",
+            "key": row["line"],
+            "error": row["error"],
+        }
+        for row in parse_errors
+    )
+    if listing_plan and not listed and not errors:
+        errors.append(
+            {
+                "operation": "parse",
+                "bucket": "",
+                "key": str(raw_output),
+                "error": "collector returned no sized objects",
+            }
+        )
+    if result.returncode != 0 or errors:
+        _write_csv(
+            phase / "inventory-errors.csv",
+            errors,
+            ["operation", "bucket", "key", "error"],
+        )
+        if result.stderr:
+            _write_text(phase / "collector-error.txt", result.stderr)
+        _inventory_status(
+            1,
+            f"FAILED: s5cmd exit {result.returncode}; {len(errors):,} listing errors. "
+            f"Inspect {phase / 'inventory-errors.csv'} and {raw_output}",
+        )
+        raise PreparationError(
+            f"S3 listing failed; inspect {phase / 'inventory-errors.csv'}"
+        )
+    _inventory_status(
+        1,
+        f"Complete: {len(listed):,} unique objects parsed in "
+        f"{result.elapsed_seconds:,.1f}s",
+    )
+    summary = _finalize_inventory(
+        build,
+        phase,
+        listed,
+        client=client,
+        max_workers=max_workers,
+        status=_inventory_status,
+    )
+    _inventory_status(
+        4,
+        f"PASS: {_human_token_count(summary['source_uint32_values'])} source "
+        f"→ {_human_token_count(summary['target_uint32_values'])} target; "
+        f"{summary['source_family_count']:,} source families, "
+        f"{summary['subcategory_count']:,} subcategories, "
+        f"{summary['category_count']:,} categories, "
+        f"{summary['lower_group_count']:,} lower groups.",
+    )
+
+
+def _walk_json(value: Any) -> Iterator[dict[str, Any]]:
+    if isinstance(value, dict):
+        yield value
+        for child in value.values():
+            yield from _walk_json(child)
+    elif isinstance(value, list):
+        for child in value:
+            yield from _walk_json(child)
+    elif isinstance(value, str) and value.startswith("{"):
+        try:
+            yield from _walk_json(json.loads(value))
+        except json.JSONDecodeError:
+            return
+
+
+def _first_value(nodes: Sequence[dict[str, Any]], names: set[str]) -> Any:
+    normalized = {re.sub(r"[^a-z0-9]", "", name.lower()) for name in names}
+    for node in nodes:
+        for key, value in node.items():
+            if re.sub(r"[^a-z0-9]", "", key.lower()) in normalized and value not in {
+                None,
+                "",
+            }:
+                return value
+    return None
+
+
+def _parse_s5cmd_jsonl(
+    path: Path,
+) -> tuple[dict[tuple[str, str], S3Object], list[dict[str, str]]]:
+    objects: dict[tuple[str, str], S3Object] = {}
+    errors: list[dict[str, str]] = []
+    # Records naming an object with no parseable size are counted and reported
+    # as a warning, not treated as a collection failure.
+    unsized_records = 0
+    with path.open(encoding="utf-8") as f:
+        for line_number, line in enumerate(f, start=1):
+            if not line.strip():
+                continue
+            try:
+                record = json.loads(line)
+            except json.JSONDecodeError as exc:
+                errors.append(
+                    {"line": str(line_number), "error": f"invalid JSON: {exc}"}
+                )
+                continue
+            if record.get("error"):
+                errors.append({"line": str(line_number), "error": str(record["error"])})
+                continue
+            nodes = list(_walk_json(record))
+            uri = _first_value(nodes, {"source", "url", "uri", "path"})
+            key_value = _first_value(nodes, {"key", "name"})
+            bucket_value = _first_value(nodes, {"bucket"})
+            if isinstance(key_value, str) and key_value.startswith("s3://"):
+                uri = key_value
+            elif (
+                not (isinstance(uri, str) and uri.startswith("s3://"))
+                and bucket_value
+                and key_value
+            ):
+                uri = f"s3://{bucket_value}/{key_value}"
+            if (
+                not isinstance(uri, str)
+                or not uri.startswith("s3://")
+                or any(c in uri for c in "*?[")
+            ):
+                continue
+            size = _first_value(
+                nodes, {"size", "size_bytes", "content_length", "contentlength"}
+            )
+            try:
+                size_int = int(size)
+            except (TypeError, ValueError):
+                unsized_records += 1
+                continue
+            parsed = urlparse(uri)
+            key = parsed.path.lstrip("/")
+            objects[(parsed.netloc, key)] = S3Object(
+                bucket=parsed.netloc,
+                key=key,
+                size_bytes=size_int,
+                etag=str(_first_value(nodes, {"etag", "e_tag"}) or "").strip('"'),
+                last_modified=str(
+                    _first_value(nodes, {"last_modified", "lastmodified", "modtime"})
+                    or ""
+                ),
+                storage_class=str(
+                    _first_value(nodes, {"storage_class", "storageclass"}) or ""
+                ),
+                source="s5cmd",
+            )
+    if unsized_records:
+        print(
+            f"WARNING: ignored {unsized_records:,} collector record(s) in {path} "
+            "that named an object with no parseable size",
+            flush=True,
+        )
+    return objects, errors
+
+
+def _finalize_inventory(
+    build: Path,
+    phase: Path,
+    listed: dict[tuple[str, str], S3Object],
+    client: Any,
+    max_workers: int,
+    status: Callable[[int, str], None] | None = None,
+) -> dict[str, Any]:
+    emit = status or (lambda _step, _message: None)
+    catalog_matches = _read_csv(build / "01-plan/resolution/catalog-matches.csv")
+    direct_patterns = _read_csv(build / "01-plan/resolution/direct-s3-patterns.csv")
+    normalized_paths = _read_csv(build / "01-plan/resolution/normalized-paths.csv")
+    emit(
+        2,
+        f"Resolving required objects for {len(normalized_paths):,} YAML paths",
+    )
+    membership: list[dict[str, Any]] = []
+    resolution_failures: list[dict[str, str]] = []
+
+    catalog_buckets_by_path: dict[str, set[str]] = defaultdict(set)
+    catalog_lines_by_path: dict[str, list[str]] = defaultdict(list)
+    for row in catalog_matches:
+        catalog_buckets_by_path[row["path_id"]].add(row["bucket"])
+        catalog_lines_by_path[row["path_id"]].append(row["catalog_line"])
+    normalized_paths_by_id = {row["path_id"]: row for row in normalized_paths}
+    for path_id, buckets in catalog_buckets_by_path.items():
+        path = normalized_paths_by_id[path_id]
+        key_pattern = _catalog_pattern(path["yaml_path"])
+        matched = [
+            obj
+            for (bucket, key), obj in listed.items()
+            if bucket in buckets
+            and key.endswith(".npy")
+            and _matches_key(key, key_pattern)
+        ]
+        if not matched:
+            resolution_failures.append(
+                {
+                    "path_id": path_id,
+                    "yaml_path": path["yaml_path"],
+                    "reason": "catalog-resolved YAML pattern returned no NPY objects",
+                }
+            )
+        for obj in sorted(matched, key=lambda item: (item.bucket, item.key)):
+            membership.append(
+                {
+                    **path,
+                    "bucket": obj.bucket,
+                    "key": obj.key,
+                    "catalog_line": ";".join(catalog_lines_by_path[path_id]),
+                }
+            )
+    for pattern in direct_patterns:
+        matches = [
+            obj
+            for (bucket, key), obj in listed.items()
+            if bucket == pattern["bucket"]
+            and key.endswith(".npy")
+            and _matches_key(key, pattern["key_pattern"])
+        ]
+        if not matches:
+            resolution_failures.append(
+                {
+                    "path_id": pattern["path_id"],
+                    "yaml_path": pattern["yaml_path"],
+                    "reason": "YAML pattern returned no NPY objects",
+                }
+            )
+        for obj in sorted(matches, key=lambda item: item.key):
+            membership.append(
+                {
+                    "path_id": pattern["path_id"],
+                    "leaf_id": pattern["leaf_id"],
+                    "mix_index": pattern["mix_index"],
+                    "mix_name": pattern["mix_name"],
+                    "category_index": pattern["category_index"],
+                    "category_name": pattern["category_name"],
+                    "active": pattern["active"],
+                    "yaml_path": pattern["yaml_path"],
+                    "bucket": obj.bucket,
+                    "key": obj.key,
+                    "catalog_line": "",
+                }
+            )
+
+    required_keys: set[tuple[str, str]] = set()
+    for row in membership:
+        required_keys.add((row["bucket"], row["key"]))
+        required_keys.add((row["bucket"], _pair_metadata_key(row["key"])))
+
+    missing = sorted(required_keys - set(listed))
+    emit(
+        2,
+        f"Resolved {len(membership):,} NPY memberships requiring "
+        f"{len(required_keys):,} unique NPY/metadata objects",
+    )
+    head_errors: list[dict[str, str]] = []
+    if missing:
+        head_total = len(missing)
+        progress_interval = max(1, math.ceil(head_total / 10))
+        emit(
+            2,
+            f"Checking {head_total:,} objects absent from bulk results with "
+            f"{max_workers:,} concurrent HeadObject requests",
+        )
+
+        def report_head_progress(completed: int, total: int) -> None:
+            if completed == total or completed % progress_interval == 0:
+                emit(2, f"Exact checks: {completed:,}/{total:,} complete")
+
+        _head_missing_objects(
+            client,
+            missing,
+            listed,
+            head_errors,
+            max_workers=max_workers,
+            on_progress=report_head_progress,
+        )
+        missing = sorted(required_keys - set(listed))
+    else:
+        emit(2, "All required objects were present in the bulk results")
+
+    missing_rows = [
+        {
+            "bucket": bucket,
+            "key": key,
+            "object_type": "npy" if key.endswith(".npy") else "metadata",
+        }
+        for bucket, key in missing
+    ]
+    required_rows: list[dict[str, Any]] = []
+    invalid_sizes: list[dict[str, Any]] = []
+    for row in membership:
+        npy = listed.get((row["bucket"], row["key"]))
+        metadata_key = _pair_metadata_key(row["key"])
+        metadata = listed.get((row["bucket"], metadata_key))
+        if npy is None or metadata is None:
+            continue
+        if npy.size_bytes <= 0 or npy.size_bytes % UINT32_BYTES:
+            invalid_sizes.append(
+                {
+                    "bucket": npy.bucket,
+                    "key": npy.key,
+                    "size_bytes": npy.size_bytes,
+                    "reason": "non-positive or not divisible by four",
+                }
+            )
+        required_rows.append(
+            {
+                **row,
+                "npy_uri": npy.uri,
+                "metadata_uri": metadata.uri,
+                "npy_size_bytes": npy.size_bytes,
+                "estimated_uint32_values": npy.size_bytes // UINT32_BYTES,
+                "npy_etag": npy.etag,
+                "npy_last_modified": npy.last_modified,
+                "metadata_size_bytes": metadata.size_bytes,
+                "metadata_etag": metadata.etag,
+                "inventory_source": npy.source,
+            }
+        )
+
+    all_objects = [
+        {
+            "bucket": obj.bucket,
+            "key": obj.key,
+            "size_bytes": obj.size_bytes,
+            "etag": obj.etag,
+            "last_modified": obj.last_modified,
+            "storage_class": obj.storage_class,
+            "source": obj.source,
+            "required": str((obj.bucket, obj.key) in required_keys).lower(),
+        }
+        for obj in sorted(listed.values(), key=lambda item: (item.bucket, item.key))
+    ]
+    required_fields = [
+        "path_id",
+        "leaf_id",
+        "mix_index",
+        "mix_name",
+        "category_index",
+        "category_name",
+        "active",
+        "yaml_path",
+        "bucket",
+        "key",
+        "npy_uri",
+        "metadata_uri",
+        "npy_size_bytes",
+        "estimated_uint32_values",
+        "npy_etag",
+        "npy_last_modified",
+        "metadata_size_bytes",
+        "metadata_etag",
+        "inventory_source",
+    ]
+    _write_csv(
+        phase / "normalized-s3-inventory.csv",
+        all_objects,
+        list(all_objects[0]) if all_objects else [],
+    )
+    _write_csv(phase / "required-objects.csv", required_rows, required_fields)
+    _write_csv(
+        phase / "missing-objects.csv", missing_rows, ["bucket", "key", "object_type"]
+    )
+    _write_csv(
+        phase / "invalid-npy-sizes.csv",
+        invalid_sizes,
+        ["bucket", "key", "size_bytes", "reason"],
+    )
+    _write_csv(
+        phase / "path-resolution-failures.csv",
+        resolution_failures,
+        ["path_id", "yaml_path", "reason"],
+    )
+    _write_csv(
+        phase / "head-errors.csv", head_errors, ["operation", "bucket", "key", "error"]
+    )
+    original_by_leaf: dict[str, dict[str, int]] = defaultdict(dict)
+    for row in required_rows:
+        original_by_leaf[row["leaf_id"]][row["npy_uri"]] = int(
+            row["estimated_uint32_values"]
+        )
+    normalized_mix = _read_csv(build / "01-plan/resolution/normalized-mix.csv")
+    original_total = sum(sum(objects.values()) for objects in original_by_leaf.values())
+    target_total = sum(int(row["target_uint32_values"]) for row in normalized_mix)
+    maximum_expected_upsample_rate = _load_build(build)["settings"].get(
+        "maximum_expected_upsample_rate"
+    )
+    sampling_rate_rows: list[dict[str, Any]] = []
+    sampling_rate_limit_failures = 0
+    for row in normalized_mix:
+        if row["active"] != "true":
+            continue
+        original = sum(original_by_leaf[row["leaf_id"]].values())
+        target = int(row["target_uint32_values"])
+        sample_rate = target / original if original else None
+        exceeds_limit = (
+            maximum_expected_upsample_rate is not None
+            and sample_rate is not None
+            and sample_rate > float(maximum_expected_upsample_rate)
+        )
+        sampling_rate_limit_failures += int(exceeds_limit)
+        sampling_rate_rows.append(
+            {
+                "leaf_id": row["leaf_id"],
+                "mix_name": row["mix_name"],
+                "category_name": row["category_name"],
+                "original_uint32_values": original,
+                "target_uint32_values": target,
+                "sample_rate": f"{sample_rate:.12g}" if sample_rate is not None else "",
+                "maximum_expected_upsample_rate": (
+                    maximum_expected_upsample_rate
+                    if maximum_expected_upsample_rate is not None
+                    else ""
+                ),
+                "status": (
+                    "above_expected_maximum"
+                    if exceeds_limit
+                    else "within_expected_range"
+                    if maximum_expected_upsample_rate is not None
+                    else "not_checked"
+                ),
+            }
+        )
+    sampling_rate_rows.sort(
+        key=lambda row: (float(row["sample_rate"] or "inf"), row["leaf_id"]),
+        reverse=True,
+    )
+    _write_csv(
+        phase / "sampling-rate-audit.csv",
+        sampling_rate_rows,
+        [
+            "leaf_id",
+            "mix_name",
+            "category_name",
+            "original_uint32_values",
+            "target_uint32_values",
+            "sample_rate",
+            "maximum_expected_upsample_rate",
+            "status",
+        ],
+    )
+    _, _, sampling_ratio = _sampling_change(original_total, target_total)
+    sampling_rate, _ = _sampling_rate_label(original_total, target_total)
+    summary = {
+        "created_at": _utc_now(),
+        "listed_objects": len(listed),
+        "required_membership_rows": len(required_rows),
+        "unique_required_objects_including_metadata": len(required_keys),
+        "missing_objects": len(missing),
+        "path_resolution_failures": len(resolution_failures),
+        "invalid_npy_sizes": len(invalid_sizes),
+        "head_errors": len(head_errors),
+        "sampling_rate_limit_failures": sampling_rate_limit_failures,
+        "maximum_expected_upsample_rate": maximum_expected_upsample_rate,
+        "source_uint32_values": original_total,
+        "original_uint32_values": original_total,
+        "target_uint32_values": target_total,
+        "token_delta": target_total - original_total,
+        "sampling_ratio": sampling_ratio,
+        "sampling_rate": sampling_rate,
+    }
+    validation_details = (
+        f"Validation: {len(required_rows):,} NPY memberships, "
+        f"{len(missing):,} missing objects, "
+        f"{len(resolution_failures):,} unresolved YAML patterns, "
+        f"{len(invalid_sizes):,} invalid NPY sizes, {len(head_errors):,} HEAD errors"
+    )
+    if maximum_expected_upsample_rate is not None:
+        validation_details += (
+            f", {sampling_rate_limit_failures:,} sampling-rate violations"
+        )
+    emit(3, validation_details)
+    emit(
+        3,
+        f"Estimated source tokens: {_human_token_count(original_total)} "
+        f"({original_total:,} uint32 values)",
+    )
+    emit(4, f"Finalizing inventory report and summary in {phase}")
+    detail_metadata = _render_inventory_report(
+        phase,
+        normalized_mix=normalized_mix,
+        normalized_paths=normalized_paths,
+        required_rows=required_rows,
+        all_objects=all_objects,
+        missing_rows=missing_rows,
+        resolution_failures=resolution_failures,
+        invalid_sizes=invalid_sizes,
+        sampling_rate_rows=sampling_rate_rows,
+    )
+    summary.update(detail_metadata)
+    _write_json(
+        phase / "inventory-summary.json",
+        summary,
+    )
+    if (
+        resolution_failures
+        or missing
+        or invalid_sizes
+        or head_errors
+        or sampling_rate_limit_failures
+    ):
+        emit(4, f"FAILED: inventory validation did not pass. Inspect {phase}")
+        raise PreparationError(
+            f"Inventory validation failed; inspect artifacts in {phase}"
+        )
+    return summary
+
+
+def _allocate_object_sampling(
+    target: int, sizes: Sequence[int]
+) -> tuple[list[int], list[int], int]:
+    """Apply one sampling rate to every shard, with partial quotas for the residual.
+
+    A 0.30 rate assigns 30% of every shard rather than 30% of the shard files.
+    Largest-remainder apportionment makes the planned total equal target, with
+    each per-shard quota within one uint32 value of its ideal. target and sizes
+    are uint32 value counts. Returns the per-shard repetition counts, the
+    partial uint32 targets, and the planned total.
+    """
+
+    if target <= 0 or not sizes or any(size <= 0 for size in sizes):
+        raise ValueError(
+            "Allocation requires a positive target and positive object sizes"
+        )
+    available = sum(sizes)
+    base = target // available
+    repetitions = [base for _ in sizes]
+    residual = target - base * available
+    partial_targets: list[int] = []
+    remainders: list[tuple[int, int]] = []
+    for index, size in enumerate(sizes):
+        partial, remainder = divmod(residual * size, available)
+        partial_targets.append(partial)
+        remainders.append((remainder, index))
+    undistributed = residual - sum(partial_targets)
+    for _, index in sorted(remainders, key=lambda item: (-item[0], item[1]))[
+        :undistributed
+    ]:
+        partial_targets[index] += 1
+
+    for index, size in enumerate(sizes):
+        if partial_targets[index] == size:
+            repetitions[index] += 1
+            partial_targets[index] = 0
+        elif partial_targets[index] > size:
+            raise AssertionError("Proportional partial target exceeds its source")
+        residual_allocation = (repetitions[index] - base) * size + partial_targets[
+            index
+        ]
+        if abs(residual_allocation * available - residual * size) >= available:
+            raise AssertionError("Per-shard sampling quota is not proportional")
+        if residual * size >= available and residual_allocation <= 0:
+            raise AssertionError(
+                "A shard with a positive proportional quota was dropped"
+            )
+    planned = sum(
+        size * repeat + partial
+        for size, repeat, partial in zip(sizes, repetitions, partial_targets)
+    )
+    if planned != target:
+        raise AssertionError(
+            f"Allocation did not preserve target: {planned} != {target}"
+        )
+    return repetitions, partial_targets, planned
+
+
+def _execution_unit_sizes(rows: Sequence[dict[str, Any]]) -> dict[str, int]:
+    def partial_metadata_bytes(row: dict[str, Any]) -> int:
+        partial_target = int(row.get("partial_target_uint32_values", 0))
+        if not partial_target:
+            return 0
+        return math.ceil(
+            int(row["metadata_size_bytes"])
+            * partial_target
+            / int(row["estimated_uint32_values"])
+        )
+
+    input_npy_bytes = sum(int(row["npy_size_bytes"]) for row in rows)
+    input_metadata_bytes = sum(int(row["metadata_size_bytes"]) for row in rows)
+    output_npy_bytes = sum(
+        int(row["npy_size_bytes"]) * int(row["repeat_count"])
+        + int(row.get("partial_target_uint32_values", 0)) * UINT32_BYTES
+        for row in rows
+    )
+    estimated_output_metadata_bytes = sum(
+        int(row["metadata_size_bytes"]) * int(row["repeat_count"])
+        + partial_metadata_bytes(row)
+        for row in rows
+    )
+    estimated_selection_index_bytes = sum(partial_metadata_bytes(row) for row in rows)
+    return {
+        "input_npy_bytes": input_npy_bytes,
+        "input_metadata_bytes": input_metadata_bytes,
+        "output_npy_bytes": output_npy_bytes,
+        "estimated_output_metadata_bytes": estimated_output_metadata_bytes,
+        "estimated_selection_index_bytes": estimated_selection_index_bytes,
+        "estimated_peak_local_bytes": input_npy_bytes
+        + input_metadata_bytes
+        + output_npy_bytes
+        + estimated_output_metadata_bytes
+        + estimated_selection_index_bytes,
+    }
+
+
+def _execution_unit_work(rows: Sequence[dict[str, Any]], document_selection_work_passes: int) -> dict[str, int]:
+    """Estimate one execution unit's CPU work, in uint32 values.
+
+    Every planned output value is merged once. A source shard using document
+    selection also costs document_selection_work_passes full scans of its whole
+    source.
+    """
+
+    if document_selection_work_passes < 0:
+        raise ValueError("document_selection_work_passes cannot be negative")
+    planned_values = sum(
+        int(row["estimated_uint32_values"]) * int(row["repeat_count"])
+        + int(row.get("partial_target_uint32_values", 0))
+        for row in rows
+    )
+    selection_source_values = sum(
+        int(row["estimated_uint32_values"]) for row in rows if int(row.get("partial_target_uint32_values", 0)) > 0
+    )
+    return {
+        "planned_uint32_values": planned_values,
+        "document_selection_source_uint32_values": selection_source_values,
+        "estimated_work_uint32_values": planned_values + document_selection_work_passes * selection_source_values,
+    }
+
+
+def _planned_output_shards(
+    *,
+    output_npy_bytes: int,
+    input_view_count: int,
+    target_output_shard_bytes: int,
+    max_output_shards_per_unit: int,
+) -> int:
+    """Return one execution unit's output shard count.
+
+    output_npy_bytes divided by target_output_shard_bytes, clamped to at least
+    one and to at most max_output_shards_per_unit and input_view_count.
+    """
+
+    if output_npy_bytes <= 0 or input_view_count <= 0:
+        raise ValueError("Output bytes and input view count must be positive")
+    if target_output_shard_bytes <= 0 or max_output_shards_per_unit <= 0:
+        raise ValueError("Output shard target and limit must be positive")
+    size_based_count = math.ceil(output_npy_bytes / target_output_shard_bytes)
+    return min(max(1, size_based_count), max_output_shards_per_unit, input_view_count)
+
+
+def _select_worker_instance(
+    grid: Sequence[dict[str, Any]],
+    *,
+    estimated_peak_local_bytes: int,
+    estimated_work_uint32_values: int,
+    disk_headroom_ratio: float,
+) -> dict[str, Any]:
+    """Return the smallest grid entry meeting the disk and work estimates.
+
+    The returned copy adds required_local_nvme_bytes and storage_layout. Raises
+    PreparationError when no entry fits.
+    """
+
+    if estimated_peak_local_bytes <= 0 or estimated_work_uint32_values <= 0:
+        raise ValueError("Execution-unit disk and work estimates must be positive")
+    if disk_headroom_ratio <= 1:
+        raise ValueError("disk_headroom_ratio must be greater than one")
+    required_local_bytes = math.ceil(estimated_peak_local_bytes * disk_headroom_ratio)
+    for raw in grid:
+        work_limit = raw.get("max_estimated_work_uint32_values")
+        if int(raw["local_nvme_bytes"]) < required_local_bytes:
+            continue
+        if work_limit is not None and estimated_work_uint32_values > int(work_limit):
+            continue
+        selected = dict(raw)
+        selected["required_local_nvme_bytes"] = required_local_bytes
+        selected["storage_layout"] = "single" if int(raw["local_nvme_devices"]) == 1 else "raid0"
+        return selected
+    largest = grid[-1]
+    raise PreparationError(
+        "No worker instance in the configured grid can run an execution unit with "
+        f"{estimated_peak_local_bytes:,} estimated peak bytes and "
+        f"{estimated_work_uint32_values:,} estimated work values. Largest candidate "
+        f"is {largest['instance_type']} with {int(largest['local_nvme_bytes']):,} "
+        "local NVMe bytes."
+    )
+
+
+def _partition_object_uses(
+    rows: Sequence[dict[str, Any]], max_unit_working_bytes: int
+) -> list[list[dict[str, Any]]]:
+    """Split one category's object uses into units of at most max_unit_working_bytes.
+
+    Uses are ordered by npy_uri, so the partition is deterministic. A single
+    object that cannot fit with one output copy raises PreparationError.
+    """
+
+    if max_unit_working_bytes <= 0:
+        raise ValueError("max_unit_working_bytes must be positive")
+    positive = [
+        dict(row)
+        for row in rows
+        if int(row["repeat_count"]) > 0
+        or int(row.get("partial_target_uint32_values", 0)) > 0
+    ]
+    if not positive:
+        raise ValueError("An execution-unit partition requires a positive object use")
+
+    units: list[list[dict[str, Any]]] = []
+    current: list[dict[str, Any]] = []
+
+    def flush() -> None:
+        nonlocal current
+        if current:
+            units.append(current)
+            current = []
+
+    for row in sorted(positive, key=lambda item: item["npy_uri"]):
+        remaining = int(row["repeat_count"])
+        partial_target = int(row.get("partial_target_uint32_values", 0))
+        input_bytes = int(row["npy_size_bytes"]) + int(row["metadata_size_bytes"])
+        output_bytes_per_repeat = input_bytes
+        maximum_repeats_alone = (
+            max_unit_working_bytes - input_bytes
+        ) // output_bytes_per_repeat
+        if maximum_repeats_alone < 1:
+            raise PreparationError(
+                "One source object cannot fit in an execution unit with one output copy: "
+                f"{row['npy_uri']} requires at least {input_bytes + output_bytes_per_repeat} bytes, "
+                f"limit is {max_unit_working_bytes}"
+            )
+
+        whole_row = dict(row)
+        whole_row["repeat_count"] = remaining
+        if (
+            _execution_unit_sizes([whole_row])["estimated_peak_local_bytes"]
+            <= max_unit_working_bytes
+        ):
+            candidate = [*current, whole_row]
+            if (
+                current
+                and _execution_unit_sizes(candidate)["estimated_peak_local_bytes"]
+                > max_unit_working_bytes
+            ):
+                flush()
+            current.append(whole_row)
+            continue
+
+        flush()
+        while remaining or partial_target:
+            chunk = dict(row)
+            chunk["repeat_count"] = 0
+            chunk["partial_target_uint32_values"] = 0
+            if partial_target:
+                chunk["partial_target_uint32_values"] = partial_target
+                partial_target = 0
+            partial_sizes = _execution_unit_sizes([chunk])
+            if partial_sizes["estimated_peak_local_bytes"] > max_unit_working_bytes:
+                raise PreparationError(
+                    "One source object's partial-document selection cannot fit in "
+                    f"an execution unit: {row['npy_uri']} requires at least "
+                    f"{partial_sizes['estimated_peak_local_bytes']} bytes, limit is "
+                    f"{max_unit_working_bytes}"
+                )
+            remaining_bytes = (
+                max_unit_working_bytes - partial_sizes["estimated_peak_local_bytes"]
+            )
+            repeat_count = min(
+                remaining,
+                max(0, remaining_bytes // output_bytes_per_repeat),
+            )
+            if repeat_count == 0 and not chunk["partial_target_uint32_values"]:
+                repeat_count = min(remaining, maximum_repeats_alone)
+            chunk["repeat_count"] = repeat_count
+            current.append(chunk)
+            remaining -= repeat_count
+            if remaining or partial_target:
+                flush()
+
+    flush()
+    return units
+
+
+def _self_contained_launcher(
+    unit_id: str,
+    config_name: str,
+    config_text: str,
+    manifest_name: str,
+    manifest_text: str,
+) -> str:
+    config_payload = base64.b64encode(config_text.encode()).decode("ascii")
+    manifest_payload = base64.b64encode(manifest_text.encode()).decode("ascii")
+    return f"""#!/usr/bin/env bash
+set -euo pipefail
+
+export PYTHONSAFEPATH=1
+cd /tmp
+
+python_bin="${{DOLMA_PYTHON:-$HOME/.venv/bin/python}}"
+if [[ ! -x "$python_bin" ]]; then
+  python_bin=$(command -v python3.12 || command -v python3 || command -v python)
+fi
+
+status_root="${{DOLMA_STATUS_ROOT:-$HOME/dolma3p5-resharding-status}}"
+mkdir -p "$status_root"
+status_path="$status_root/{unit_id}.status"
+log_path="$status_root/{unit_id}.log"
+exec > >(tee -a "$log_path") 2>&1
+
+unit_root=$(mktemp -d "${{TMPDIR:-/tmp}}/dolma3p5-{unit_id}.XXXXXX")
+finish() {{
+  exit_code=$?
+  rm -rf -- "$unit_root"
+  if [[ $exit_code -eq 0 ]]; then
+    printf 'succeeded\n' > "$status_path"
+  else
+    printf 'failed %s\n' "$exit_code" > "$status_path"
+  fi
+  trap - EXIT
+  exit "$exit_code"
+}}
+trap finish EXIT
+printf 'running\n' > "$status_path"
+mkdir -p "$unit_root/config" "$unit_root/manifests"
+
+"$python_bin" -P - <<'PY'
+from dolma.tokenizer.reshard import RESHARDING_MANIFEST_SCHEMA_VERSION
+
+if RESHARDING_MANIFEST_SCHEMA_VERSION != {EXPECTED_MANIFEST_SCHEMA_VERSION}:
+    raise RuntimeError(
+        "Worker Dolma runtime does not match the expected manifest-resharding schema"
+    )
+PY
+
+"$python_bin" -P - "$unit_root/config/{config_name}" "$unit_root/manifests/{manifest_name}" <<'PY'
+import base64
+import pathlib
+import sys
+
+pathlib.Path(sys.argv[1]).write_bytes(base64.b64decode("{config_payload}", validate=True))
+pathlib.Path(sys.argv[2]).write_bytes(base64.b64decode("{manifest_payload}", validate=True))
+PY
+
+"$python_bin" -P -m dolma.tokenizer.reshard "$unit_root/config/{config_name}"
+"""
+
+
+def _validate_destination_root(destination_root: str) -> str:
+    parsed = urlparse(destination_root)
+    if parsed.scheme != "s3" or not parsed.netloc or parsed.query or parsed.fragment:
+        raise PreparationError("destination-root must be an s3:// URI")
+    prefix = parsed.path.strip("/")
+    components = prefix.split("/")
+    if len(components) < 2:
+        raise PreparationError(
+            "destination-root must contain at least two path components below the bucket"
+        )
+    if any(component in {"", ".", ".."} for component in components):
+        raise PreparationError(
+            "destination-root cannot contain empty, '.' or '..' components"
+        )
+    if not re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9._/-]*", prefix):
+        raise PreparationError(
+            "destination-root must use only letters, digits, '.', '_', '-', and '/'"
+        )
+    return f"s3://{parsed.netloc}/{prefix}"
+
+
+def _source_relative_directory(source_key: str) -> str:
+    """Return a source directory relative to its top-level storage prefix."""
+
+    components = tuple(source_key.split("/"))
+    if len(components) < 3 or any(
+        component in {"", ".", ".."} for component in components
+    ):
+        raise PreparationError(f"Source object has an unsafe path: {source_key}")
+    directory = components[1:-1]
+    return "/".join(directory)
+
+
+def _source_root_uri(bucket: str, source_key: str) -> str:
+    top_level = source_key.split("/", 1)[0]
+    if not bucket or not top_level or top_level in {".", ".."}:
+        raise PreparationError(
+            f"Cannot derive source root for s3://{bucket}/{source_key}"
+        )
+    return f"s3://{bucket}/{top_level}"
+
+
+def _common_path_prefix(paths: Sequence[tuple[str, ...]]) -> tuple[str, ...]:
+    common: list[str] = []
+    for components in zip(*paths):
+        if len(set(components)) != 1:
+            break
+        common.append(components[0])
+    return tuple(common)
+
+
+def _common_path_suffix(
+    paths: Sequence[tuple[str, ...]],
+    prefix_length: int,
+) -> tuple[str, ...]:
+    common_reversed: list[str] = []
+    available = min(len(path) - prefix_length for path in paths)
+    for offset in range(1, available + 1):
+        components = {path[-offset] for path in paths}
+        if len(components) != 1:
+            break
+        common_reversed.append(components.pop())
+    return tuple(reversed(common_reversed))
+
+
+def _category_output_directory(
+    objects: Sequence[dict[str, str]],
+    category_name: str,
+) -> str:
+    """Return one output directory for a category, derived from its source paths.
+
+    A single source directory is reused unchanged. Otherwise the result is the
+    sources' common prefix, the category name, and their common suffix.
+    """
+
+    if not category_name or category_name in {".", ".."} or "/" in category_name:
+        raise PreparationError(
+            f"Category name cannot be used in an output path: {category_name!r}"
+        )
+    directories = sorted(
+        {tuple(_source_relative_directory(row["key"]).split("/")) for row in objects}
+    )
+    if len(directories) == 1:
+        return "/".join(directories[0])
+    common_prefix = _common_path_prefix(directories)
+    if not common_prefix:
+        raise PreparationError(
+            f"Category {category_name!r} has source paths without a common output prefix"
+        )
+    common_suffix = _common_path_suffix(directories, len(common_prefix))
+    return "/".join((*common_prefix, category_name, *common_suffix))
+
+
+EXECUTION_MANIFEST_FIELDS = [
+    "npy_uri",
+    "metadata_uri",
+    "repeat_count",
+    "partial_target_uint32_values",
+    "selection_seed",
+    "selection_algorithm",
+    "npy_size_bytes",
+    "estimated_uint32_values",
+    "npy_etag",
+    "metadata_size_bytes",
+    "metadata_etag",
+    "leaf_id",
+    "mix_name",
+    "category_name",
+    "source_directory",
+]
+
+
+@dataclass(frozen=True)
+class _ExecutionPlanContext:
+    """Per-run inputs shared by every emitted execution unit."""
+
+    build: Path
+    build_id: str
+    settings: dict[str, Any]
+    worker_grid: list[dict[str, Any]]
+    worker_disk_headroom_ratio: float
+    document_selection_work_passes: int
+    target_output_shard_bytes: int
+    max_output_shards_per_unit: int
+    max_unit_working_bytes: int
+    requested_max_unit_working_bytes: int
+    dataset_root: str
+    local_temp_root: Path
+    configs_dir: Path
+    manifests_dir: Path
+    launcher_dir: Path
+
+
+def _require_clean_inventory(inventory_phase: Path) -> dict[str, Any]:
+    """Return inventory-summary.json, raising if it recorded a blocking failure."""
+
+    summary_path = inventory_phase / "inventory-summary.json"
+    if not summary_path.is_file():
+        raise PreparationError("Inventory is missing; rerun scripts/dolma3p5_resharding/plan.py")
+    with summary_path.open() as f:
+        summary = json.load(f)
+    blocking = sum(
+        int(summary.get(name, 0))
+        for name in (
+            "missing_objects",
+            "path_resolution_failures",
+            "direct_resolution_failures",
+            "invalid_npy_sizes",
+            "head_errors",
+            "sampling_rate_limit_failures",
+        )
+    )
+    if blocking:
+        raise PreparationError(
+            "Inventory contains blocking validation failures; inspect "
+            f"{inventory_phase / 'inventory-summary.json'} and "
+            f"{inventory_phase / 'sampling-rate-audit.csv'}"
+        )
+    return summary
+
+
+def _index_active_inventory(
+    inventory: Sequence[dict[str, str]], phase: Path
+) -> tuple[dict[str, dict[tuple[str, str], dict[str, str]]], list[str]]:
+    """Group active inventory rows by category, rejecting cross-category reuse.
+
+    Writes cross-leaf-overlaps.csv. Returns the per-category object index and
+    the distinct source storage roots.
+    """
+
+    by_leaf: dict[str, dict[tuple[str, str], dict[str, str]]] = defaultdict(dict)
+    memberships: dict[tuple[str, str], set[str]] = defaultdict(set)
+    for row in inventory:
+        if row["active"] != "true":
+            continue
+        object_id = (row["bucket"], row["key"])
+        by_leaf[row["leaf_id"]][object_id] = row
+        memberships[object_id].add(row["leaf_id"])
+    source_roots = sorted(
+        {_source_root_uri(row["bucket"], row["key"]) for row in inventory if row["active"] == "true"}
+    )
+    overlaps = [
+        {"bucket": bucket, "key": key, "leaf_ids": ";".join(sorted(leaves))}
+        for (bucket, key), leaves in memberships.items()
+        if len(leaves) > 1
+    ]
+    _write_csv(phase / "cross-leaf-overlaps.csv", overlaps, ["bucket", "key", "leaf_ids"])
+    if overlaps:
+        raise PreparationError(
+            f"Found {len(overlaps)} exact NPY object(s) assigned to multiple active categories; inspect cross-leaf-overlaps.csv"
+        )
+    return by_leaf, source_roots
+
+
+def _allocate_category_object_uses(
+    leaf: dict[str, str],
+    objects: Sequence[dict[str, str]],
+    settings: dict[str, Any],
+) -> tuple[dict[str, Any], list[dict[str, Any]], int]:
+    """Spread one category's target across its inventoried source objects.
+
+    Returns the category's allocation summary row, one planned-use row per
+    source object, and the planned uint32 total.
+    """
+
+    mix_index = int(leaf["mix_index"])
+    category_index = int(leaf["category_index"])
+    mix_name = leaf["mix_name"]
+    category_name = leaf["category_name"]
+    sizes = [int(row["estimated_uint32_values"]) for row in objects]
+    target = int(leaf["target_uint32_values"])
+    repetitions, partial_targets, planned = _allocate_object_sampling(target, sizes)
+    available = sum(sizes)
+    allocation_row = {
+        "leaf_id": leaf["leaf_id"],
+        "mix_index": mix_index,
+        "mix_name": mix_name,
+        "category_index": category_index,
+        "category_name": category_name,
+        "effective_weight": leaf["effective_weight"],
+        "target_uint32_values": target,
+        "available_uint32_values": available,
+        "planned_uint32_values": planned,
+        "token_change_from_original": planned - available,
+        "token_change_percent_from_original": f"{(planned - available) / available:.12g}",
+        "ideal_sample_rate": f"{target / available:.12g}",
+        "effective_sample_rate": f"{planned / available:.12g}",
+        "target_residual_uint32_values": planned - target,
+        "target_residual_bps_of_total": f"{10_000 * (planned - target) / int(settings['target_uint32_values']):.12g}",
+        "target_residual_fraction": f"{(planned - target) / target:.12g}",
+        "unique_object_count": len(objects),
+        "selected_object_count": sum(
+            repeat > 0 or partial > 0 for repeat, partial in zip(repetitions, partial_targets)
+        ),
+        "dropped_object_count": sum(
+            repeat == 0 and partial == 0 for repeat, partial in zip(repetitions, partial_targets)
+        ),
+        "repeated_object_count": sum(
+            size * repeat + partial > size
+            for size, repeat, partial in zip(sizes, repetitions, partial_targets)
+        ),
+        "partial_object_count": sum(value > 0 for value in partial_targets),
+        "total_object_uses": sum(repetitions) + sum(value > 0 for value in partial_targets),
+        "minimum_repetition": min(repetitions),
+        "maximum_repetition": max(repetitions),
+    }
+    object_uses: list[dict[str, Any]] = []
+    for obj, repeat_count, partial_target in zip(objects, repetitions, partial_targets):
+        selection_seed = int(settings["random_seed"]) + int(
+            hashlib.sha256(f"{leaf['leaf_id']}\0{obj['npy_uri']}".encode()).hexdigest()[:16],
+            16,
+        )
+        object_uses.append(
+            {
+                "leaf_id": leaf["leaf_id"],
+                "mix_index": mix_index,
+                "mix_name": mix_name,
+                "category_index": category_index,
+                "category_name": category_name,
+                "source_directory": obj["npy_uri"].rsplit("/", 1)[0],
+                "source_layout_prefix": _source_relative_directory(obj["key"]),
+                "path_id": obj["path_id"],
+                "lower_group": _path_subgroup(obj["yaml_path"]),
+                "yaml_path": obj["yaml_path"],
+                "npy_uri": obj["npy_uri"],
+                "metadata_uri": obj["metadata_uri"],
+                "npy_size_bytes": obj["npy_size_bytes"],
+                "estimated_uint32_values": obj["estimated_uint32_values"],
+                "npy_etag": obj["npy_etag"],
+                "metadata_size_bytes": obj["metadata_size_bytes"],
+                "metadata_etag": obj["metadata_etag"],
+                "repeat_count": repeat_count,
+                "partial_target_uint32_values": partial_target,
+                "selection_seed": selection_seed,
+                "selection_algorithm": DOCUMENT_SELECTION_ALGORITHM,
+                "planned_uint32_values": int(obj["estimated_uint32_values"]) * repeat_count + partial_target,
+            }
+        )
+    return allocation_row, object_uses, planned
+
+
+def _emit_execution_unit(
+    context: _ExecutionPlanContext,
+    *,
+    unit_id: str,
+    unit_rows: Sequence[dict[str, Any]],
+    unit_target: int,
+    unit_index: int,
+    unit_count: int,
+    leaf: dict[str, str],
+    source_layout_prefix: str,
+    category_target: int,
+    category_planned: int,
+) -> tuple[dict[str, Any], str]:
+    """Write one execution unit's manifest, reshard config, and launcher.
+
+    Writes manifests/<unit_id>.csv, config/<unit_id>.yaml, and the executable
+    launcher-scripts/<unit_id>.sh. Returns the unit's config-index row and its
+    single-unit debugging command.
+    """
+
+    settings = context.settings
+    unit_sizes = _execution_unit_sizes(unit_rows)
+    unit_work = _execution_unit_work(unit_rows, context.document_selection_work_passes)
+    worker = _select_worker_instance(
+        context.worker_grid,
+        estimated_peak_local_bytes=unit_sizes["estimated_peak_local_bytes"],
+        estimated_work_uint32_values=unit_work["estimated_work_uint32_values"],
+        disk_headroom_ratio=context.worker_disk_headroom_ratio,
+    )
+    unit_planned = unit_sizes["output_npy_bytes"] // UINT32_BYTES
+    unit_max_repeat = max(int(row["repeat_count"]) for row in unit_rows)
+    unit_partial_objects = sum(int(row.get("partial_target_uint32_values", 0)) > 0 for row in unit_rows)
+    input_view_count = sum(
+        int(row["repeat_count"]) + (int(row.get("partial_target_uint32_values", 0)) > 0)
+        for row in unit_rows
+    )
+    manifest_path = context.manifests_dir / f"{unit_id}.csv"
+    _write_csv(manifest_path, unit_rows, EXECUTION_MANIFEST_FIELDS)
+    max_num_files = _planned_output_shards(
+        output_npy_bytes=unit_sizes["output_npy_bytes"],
+        input_view_count=input_view_count,
+        target_output_shard_bytes=context.target_output_shard_bytes,
+        max_output_shards_per_unit=context.max_output_shards_per_unit,
+    )
+    max_workers = min(int(settings["max_workers_per_reshard"]), int(worker["vcpus"]))
+    destination_index = f"{unit_index:0{EXECUTION_UNIT_INDEX_WIDTH}d}"
+    destination = f"{context.dataset_root}/{source_layout_prefix}/{destination_index}"
+    config = {
+        "destination_prefix": destination,
+        "source_manifests": [{"manifest": f"../manifests/{manifest_path.name}"}],
+        "local_tempdir": str(context.local_temp_root / context.build_id / unit_id),
+        "max_num_files": max_num_files,
+        "max_workers": max_workers,
+        "s5cmd_download_concurrency": int(settings["s5cmd_download_concurrency"]),
+        "random_seed": int(settings["random_seed"])
+        + int(hashlib.sha256(unit_id.encode()).hexdigest()[:16], 16),
+        "tokenizer_name_or_path": str(settings["tokenizer_name_or_path"]),
+        "allow_existing_destination": False,
+    }
+    config_path = context.configs_dir / f"{unit_id}.yaml"
+    config_text = yaml.safe_dump(config, sort_keys=False)
+    _write_text(config_path, config_text)
+    launcher_path = context.launcher_dir / f"{unit_id}.sh"
+    _write_text(
+        launcher_path,
+        _self_contained_launcher(
+            unit_id=unit_id,
+            config_name=config_path.name,
+            config_text=config_text,
+            manifest_name=manifest_path.name,
+            manifest_text=manifest_path.read_text(),
+        ),
+    )
+    launcher_path.chmod(0o755)
+    unit_row = {
+        "unit_id": unit_id,
+        "leaf_id": leaf["leaf_id"],
+        "mix_index": int(leaf["mix_index"]),
+        "mix_name": leaf["mix_name"],
+        "category_index": int(leaf["category_index"]),
+        "category_name": leaf["category_name"],
+        "source_directories": ";".join(sorted({row["source_directory"] for row in unit_rows})),
+        "source_directory_count": len({row["source_directory"] for row in unit_rows}),
+        "source_layout_prefix": source_layout_prefix,
+        "destination_index": destination_index,
+        "unit_index": unit_index + 1,
+        "unit_count_for_category": unit_count,
+        "config_path": str(config_path.relative_to(context.build)),
+        "manifest_path": str(manifest_path.relative_to(context.build)),
+        "launcher_path": str(launcher_path.relative_to(context.build)),
+        "destination_prefix": destination,
+        "target_uint32_values": unit_target,
+        "planned_uint32_values": unit_planned,
+        "target_residual_uint32_values": unit_planned - unit_target,
+        "category_target_uint32_values": category_target,
+        "category_planned_uint32_values": category_planned,
+        "input_npy_bytes": unit_sizes["input_npy_bytes"],
+        "input_metadata_bytes": unit_sizes["input_metadata_bytes"],
+        "output_npy_bytes": unit_sizes["output_npy_bytes"],
+        "estimated_output_metadata_bytes": unit_sizes["estimated_output_metadata_bytes"],
+        "estimated_selection_index_bytes": unit_sizes["estimated_selection_index_bytes"],
+        "estimated_peak_local_bytes": unit_sizes["estimated_peak_local_bytes"],
+        "max_unit_working_bytes": context.max_unit_working_bytes,
+        "requested_max_unit_working_bytes": context.requested_max_unit_working_bytes,
+        "working_budget_utilization": f"{unit_sizes['estimated_peak_local_bytes'] / context.max_unit_working_bytes:.12g}",
+        "document_selection_source_uint32_values": unit_work["document_selection_source_uint32_values"],
+        "document_selection_work_passes": context.document_selection_work_passes,
+        "estimated_work_uint32_values": unit_work["estimated_work_uint32_values"],
+        "worker_instance_type": worker["instance_type"],
+        "worker_vcpus": worker["vcpus"],
+        "worker_local_nvme_devices": worker["local_nvme_devices"],
+        "worker_local_nvme_bytes": worker["local_nvme_bytes"],
+        "worker_required_local_nvme_bytes": worker["required_local_nvme_bytes"],
+        "worker_disk_headroom_ratio": f"{context.worker_disk_headroom_ratio:.12g}",
+        "worker_storage_layout": worker["storage_layout"],
+        "worker_disk_utilization": f"{unit_sizes['estimated_peak_local_bytes'] / int(worker['local_nvme_bytes']):.12g}",
+        "unique_object_count": len(unit_rows),
+        "partial_object_count": unit_partial_objects,
+        "input_view_count": input_view_count,
+        "allowed_materialized_target_residual_uint32_values": (
+            math.ceil(unit_target * float(settings["max_materialized_unit_target_residual_fraction"]))
+            if unit_partial_objects
+            else 0
+        ),
+        "maximum_repetition": unit_max_repeat,
+        "target_output_shard_bytes": context.target_output_shard_bytes,
+        "max_output_shards_per_unit": context.max_output_shards_per_unit,
+        "planned_output_shard_count": max_num_files,
+        "average_output_shard_bytes": math.ceil(unit_sizes["output_npy_bytes"] / max_num_files),
+        "max_num_files": max_num_files,
+        "max_workers": max_workers,
+    }
+    return unit_row, f"python -m dolma.tokenizer.reshard {shlex.quote(str(config_path))}"
+
+
+def _category_execution_row(
+    leaf: dict[str, str],
+    unit_rows: Sequence[dict[str, Any]],
+    *,
+    target: int,
+    planned: int,
+    max_unit_working_bytes: int,
+) -> dict[str, Any]:
+    """Summarize one category's execution units for category-execution-summary.csv."""
+
+    return {
+        "leaf_id": leaf["leaf_id"],
+        "mix_index": int(leaf["mix_index"]),
+        "mix_name": leaf["mix_name"],
+        "category_index": int(leaf["category_index"]),
+        "category_name": leaf["category_name"],
+        "target_uint32_values": target,
+        "planned_uint32_values": planned,
+        "execution_unit_count": len(unit_rows),
+        "largest_estimated_peak_local_bytes": max(
+            int(row["estimated_peak_local_bytes"]) for row in unit_rows
+        ),
+        "total_input_bytes_across_units": sum(
+            int(row["input_npy_bytes"]) + int(row["input_metadata_bytes"]) for row in unit_rows
+        ),
+        "max_unit_working_bytes": max_unit_working_bytes,
+        "worker_instance_types": ";".join(
+            sorted({str(row["worker_instance_type"]) for row in unit_rows})
+        ),
+        "planned_output_shard_count": sum(
+            int(row["planned_output_shard_count"]) for row in unit_rows
+        ),
+    }
+
+
+def propose_configs(args: argparse.Namespace) -> None:
+    """Allocate per-object sampling and partition it into worker execution units.
+
+    Writes the 01-plan/execution stage: one config, manifest, and launcher per
+    unit, the allocation and unit CSVs, execution-units.jsonl,
+    dataset-layout.json, runtime-requirements.json, dataset-prefixes.txt, and
+    proposal-summary.json. Also rewrites the combined 01-plan/report.html.
+    """
+
+    build = args.build.resolve()
+    manifest = _load_build(build)
+    inventory_phase = build / "01-plan/inventory"
+    inventory_summary = _require_clean_inventory(inventory_phase)
+
+    phase = _reset_plan_stage(build, "execution")
+    configs_dir = phase / "config"
+    manifests_dir = phase / "manifests"
+    plots_dir = phase / "plots"
+    launcher_dir = phase / "launcher-scripts"
+    configs_dir.mkdir()
+    manifests_dir.mkdir()
+    plots_dir.mkdir()
+    launcher_dir.mkdir()
+    destination_root = _validate_destination_root(args.destination_root)
+    local_temp_root = Path(args.local_temp_root)
+    if not local_temp_root.is_absolute():
+        raise PreparationError("local-temp-root must be an absolute path")
+    requested_max_unit_working_bytes = int(args.max_unit_working_bytes)
+    if requested_max_unit_working_bytes <= 0:
+        raise PreparationError("max-unit-working-bytes must be positive")
+
+    normalized_mix = _read_csv(build / "01-plan/resolution/normalized-mix.csv")
+    inventory = _read_csv(inventory_phase / "required-objects.csv")
+    by_leaf, source_roots = _index_active_inventory(inventory, phase)
+
+    settings = manifest["settings"]
+    worker_grid = _validate_worker_instance_grid(settings["worker_instance_grid"])
+    worker_disk_headroom_ratio = float(settings["worker_disk_headroom_ratio"])
+    document_selection_work_passes = int(settings["document_selection_work_passes"])
+    target_output_shard_bytes = int(settings["target_output_shard_bytes"])
+    max_output_shards_per_unit = int(settings["max_output_shards_per_unit"])
+    largest_worker = worker_grid[-1]
+    largest_worker_working_bytes = math.floor(int(largest_worker["local_nvme_bytes"]) / worker_disk_headroom_ratio)
+    max_unit_working_bytes = min(requested_max_unit_working_bytes, largest_worker_working_bytes)
+    context = _ExecutionPlanContext(
+        build=build,
+        build_id=manifest["build_id"],
+        settings=settings,
+        worker_grid=worker_grid,
+        worker_disk_headroom_ratio=worker_disk_headroom_ratio,
+        document_selection_work_passes=document_selection_work_passes,
+        target_output_shard_bytes=target_output_shard_bytes,
+        max_output_shards_per_unit=max_output_shards_per_unit,
+        max_unit_working_bytes=max_unit_working_bytes,
+        requested_max_unit_working_bytes=requested_max_unit_working_bytes,
+        dataset_root=f"{destination_root}/{manifest['build_id']}",
+        local_temp_root=local_temp_root,
+        configs_dir=configs_dir,
+        manifests_dir=manifests_dir,
+        launcher_dir=launcher_dir,
+    )
+
+    allocation_rows: list[dict[str, Any]] = []
+    object_use_rows: list[dict[str, Any]] = []
+    config_index: list[dict[str, Any]] = []
+    category_execution_rows: list[dict[str, Any]] = []
+    local_unit_commands: list[str] = []
+    next_execution_unit_index = 0
+    total_planned = 0
+    active_leaves = sorted(
+        (row for row in normalized_mix if row["active"] == "true"),
+        key=lambda row: (int(row["mix_index"]), int(row["category_index"])),
+    )
+    for leaf in active_leaves:
+        objects = sorted(
+            by_leaf.get(leaf["leaf_id"], {}).values(),
+            key=lambda row: (row["bucket"], row["key"]),
+        )
+        if not objects:
+            raise PreparationError(f"Active category has no inventoried objects: {leaf['leaf_id']}")
+        target = int(leaf["target_uint32_values"])
+        allocation_row, object_uses, planned = _allocate_category_object_uses(leaf, objects, settings)
+        allocation_rows.append(allocation_row)
+        object_use_rows.extend(object_uses)
+        total_planned += planned
+
+        selected_uses = [
+            use
+            for use in object_uses
+            if int(use["repeat_count"]) > 0 or int(use["partial_target_uint32_values"]) > 0
+        ]
+        units = _partition_object_uses(selected_uses, max_unit_working_bytes)
+        if len(units) > 10**EXECUTION_UNIT_INDEX_WIDTH:
+            raise PreparationError(
+                f"Category {leaf['leaf_id']} needs {len(units):,} execution units, which exceeds "
+                f"the {EXECUTION_UNIT_INDEX_WIDTH}-digit destination counter"
+            )
+        source_layout_prefix = _category_output_directory(objects, leaf["category_name"])
+        # A unit's target is the output its own partition plans.
+        unit_targets = [_execution_unit_sizes(unit)["output_npy_bytes"] // UINT32_BYTES for unit in units]
+
+        leaf_unit_rows: list[dict[str, Any]] = []
+        for unit_index, (unit_rows, unit_target) in enumerate(zip(units, unit_targets)):
+            if next_execution_unit_index >= 10**EXECUTION_UNIT_INDEX_WIDTH:
+                raise PreparationError(
+                    f"Execution plan exceeds the {EXECUTION_UNIT_INDEX_WIDTH}-digit unit ID counter"
+                )
+            unit_id = f"{next_execution_unit_index:0{EXECUTION_UNIT_INDEX_WIDTH}d}"
+            next_execution_unit_index += 1
+            unit_row, local_command = _emit_execution_unit(
+                context,
+                unit_id=unit_id,
+                unit_rows=unit_rows,
+                unit_target=unit_target,
+                unit_index=unit_index,
+                unit_count=len(units),
+                leaf=leaf,
+                source_layout_prefix=source_layout_prefix,
+                category_target=target,
+                category_planned=planned,
+            )
+            leaf_unit_rows.append(unit_row)
+            config_index.append(unit_row)
+            local_unit_commands.append(local_command)
+
+        category_execution_rows.append(
+            _category_execution_row(
+                leaf,
+                leaf_unit_rows,
+                target=target,
+                planned=planned,
+                max_unit_working_bytes=max_unit_working_bytes,
+            )
+        )
+
+    worker_instance_counts = dict(sorted(Counter(row["worker_instance_type"] for row in config_index).items()))
+    planned_output_shard_count = sum(int(row["planned_output_shard_count"]) for row in config_index)
+    planned_output_file_count = planned_output_shard_count * 2
+    allocation_fields = [
+        "leaf_id",
+        "mix_index",
+        "mix_name",
+        "category_index",
+        "category_name",
+        "effective_weight",
+        "target_uint32_values",
+        "available_uint32_values",
+        "planned_uint32_values",
+        "token_change_from_original",
+        "token_change_percent_from_original",
+        "ideal_sample_rate",
+        "effective_sample_rate",
+        "target_residual_uint32_values",
+        "target_residual_bps_of_total",
+        "target_residual_fraction",
+        "unique_object_count",
+        "selected_object_count",
+        "dropped_object_count",
+        "repeated_object_count",
+        "partial_object_count",
+        "total_object_uses",
+        "minimum_repetition",
+        "maximum_repetition",
+    ]
+    _write_csv(phase / "category-allocation.csv", allocation_rows, allocation_fields)
+    _write_csv(
+        phase / "planned-object-uses.csv",
+        object_use_rows,
+        list(object_use_rows[0]) if object_use_rows else [],
+    )
+    _write_csv(
+        phase / "config-index.csv",
+        config_index,
+        list(config_index[0]) if config_index else [],
+    )
+    _write_csv(
+        phase / "category-execution-summary.csv",
+        category_execution_rows,
+        list(category_execution_rows[0]) if category_execution_rows else [],
+    )
+    _write_text(
+        phase / "execution-units.jsonl",
+        "\n".join(json.dumps(row, sort_keys=True) for row in config_index) + "\n",
+    )
+    _write_text(
+        phase / "LOCAL-UNIT-COMMANDS.txt",
+        "# INERT REVIEW ARTIFACT. These commands were not run.\n"
+        "# Each command is for debugging one unit only; do not run this entire file on one machine.\n"
+        + "\n".join(local_unit_commands)
+        + "\n",
+    )
+    _write_json(
+        phase / "dataset-layout.json",
+        {
+            "schema_version": EXECUTION_LAYOUT_SCHEMA_VERSION,
+            "build_id": manifest["build_id"],
+            "dataset_root": f"{destination_root}/{manifest['build_id']}",
+            "layout": DESTINATION_LAYOUT,
+            "source_roots": source_roots,
+            "execution_unit_index_width": EXECUTION_UNIT_INDEX_WIDTH,
+            "execution_unit_id_width": EXECUTION_UNIT_INDEX_WIDTH,
+            "category_count": len(category_execution_rows),
+            "execution_unit_count": len(config_index),
+            "nominal_target_uint32_values": int(settings["target_uint32_values"]),
+            "target_uint32_values": sum(int(row["target_uint32_values"]) for row in allocation_rows),
+            "planned_uint32_values": total_planned,
+            "max_unit_working_bytes": max_unit_working_bytes,
+            "requested_max_unit_working_bytes": requested_max_unit_working_bytes,
+            "worker_disk_headroom_ratio": worker_disk_headroom_ratio,
+            "worker_instance_grid": worker_grid,
+            "worker_instance_counts": worker_instance_counts,
+            "target_output_shard_bytes": target_output_shard_bytes,
+            "max_output_shards_per_unit": max_output_shards_per_unit,
+            "planned_output_shard_count": planned_output_shard_count,
+            "planned_output_file_count": planned_output_file_count,
+            "destination_prefixes_file": "dataset-prefixes.txt",
+        },
+    )
+    _write_json(
+        phase / "runtime-requirements.json",
+        {
+            "required_python_module": "dolma.tokenizer.reshard",
+            "required_python_modules": [
+                "dolma.tokenizer.reshard",
+                "dolma.tokenizer.document_selection",
+            ],
+            "required_resharding_manifest_schema_version": EXPECTED_MANIFEST_SCHEMA_VERSION,
+            "document_selection_algorithm": DOCUMENT_SELECTION_ALGORITHM,
+            "launcher_runtime_check": True,
+        },
+    )
+    _write_text(
+        phase / "dataset-prefixes.txt",
+        "\n".join(row["destination_prefix"] for row in config_index) + "\n",
+    )
+
+    _validate_proposal(build, phase, config_index, allocation_rows, object_use_rows)
+    report_totals = _render_report(
+        build,
+        phase,
+        allocation_rows,
+        object_use_rows,
+        inventory,
+        config_index,
+        category_execution_rows,
+    )
+    inventoried_source = int(
+        inventory_summary["source_uint32_values"]
+        if "source_uint32_values" in inventory_summary
+        else inventory_summary["original_uint32_values"]
+    )
+    if report_totals["source_uint32_values"] != inventoried_source:
+        raise PreparationError(
+            "Proposal source total does not match the source inventory: "
+            f"{report_totals['source_uint32_values']:,} != {inventoried_source:,}"
+        )
+    target_residual = total_planned - report_totals["target_uint32_values"]
+    source_shards_with_document_selection = sum(
+        int(row.get("partial_target_uint32_values", 0)) > 0 for row in object_use_rows
+    )
+    _write_json(
+        phase / "proposal-summary.json",
+        {
+            "created_at": _utc_now(),
+            "build_id": manifest["build_id"],
+            "destination_root": destination_root,
+            "dataset_root": f"{destination_root}/{manifest['build_id']}",
+            "destination_layout": DESTINATION_LAYOUT,
+            "source_roots": source_roots,
+            "execution_unit_index_width": EXECUTION_UNIT_INDEX_WIDTH,
+            "execution_unit_id_width": EXECUTION_UNIT_INDEX_WIDTH,
+            "category_count": len(category_execution_rows),
+            "execution_unit_count": len(config_index),
+            "config_count": len(config_index),
+            "max_unit_working_bytes": max_unit_working_bytes,
+            "requested_max_unit_working_bytes": requested_max_unit_working_bytes,
+            "largest_estimated_peak_local_bytes": max(
+                int(row["estimated_peak_local_bytes"]) for row in config_index
+            ),
+            "worker_disk_headroom_ratio": worker_disk_headroom_ratio,
+            "document_selection_work_passes": document_selection_work_passes,
+            "worker_instance_grid": worker_grid,
+            "worker_instance_counts": worker_instance_counts,
+            "target_output_shard_bytes": target_output_shard_bytes,
+            "max_output_shards_per_unit": max_output_shards_per_unit,
+            "planned_output_shard_count": planned_output_shard_count,
+            "planned_output_file_count": planned_output_file_count,
+            "nominal_target_uint32_values": int(settings["target_uint32_values"]),
+            "target_uint32_values": report_totals["target_uint32_values"],
+            "source_uint32_values": report_totals["source_uint32_values"],
+            "planned_uint32_values": total_planned,
+            "token_change_from_source": total_planned - report_totals["source_uint32_values"],
+            "target_residual_uint32_values": target_residual,
+            "nominal_target_residual_uint32_values": total_planned - int(settings["target_uint32_values"]),
+            "source_shards_with_document_selection": source_shards_with_document_selection,
+            "document_selection_algorithm": DOCUMENT_SELECTION_ALGORITHM,
+            "max_materialized_unit_target_residual_fraction": float(
+                settings["max_materialized_unit_target_residual_fraction"]
+            ),
+            "max_materialized_total_target_residual_fraction": float(
+                settings["max_materialized_total_target_residual_fraction"]
+            ),
+            "materialization_executed": False,
+            "report_artifact": "../report.html",
+        },
+    )
+    _combine_plan_reports(build)
+    worker_summary = ", ".join(
+        f"{count:,} {instance_type}" for instance_type, count in worker_instance_counts.items()
+    )
+    print(
+        f"Execution plan ready: {_human_token_count(total_planned)} tokens across "
+        f"{len(config_index):,} execution units and {planned_output_shard_count:,} output shards; "
+        f"target residual: {target_residual:,} tokens.\n"
+        f"Planned workers: {worker_summary}.\n"
+        f"Review: {build / '01-plan/report.html'}"
+    )
+
+
+def _validate_proposal(
+    build: Path,
+    phase: Path,
+    config_index: Sequence[dict[str, Any]],
+    allocations: Sequence[dict[str, Any]],
+    object_uses: Sequence[dict[str, Any]],
+) -> None:
+    failures: list[dict[str, str]] = []
+    build_manifest = _validate_preparation_build(build)
+    worker_grid = _validate_worker_instance_grid(build_manifest["settings"]["worker_instance_grid"])
+    workers_by_type = {row["instance_type"]: row for row in worker_grid}
+    target_output_shard_bytes = int(build_manifest["settings"]["target_output_shard_bytes"])
+    max_output_shards_per_unit = int(build_manifest["settings"]["max_output_shards_per_unit"])
+    destinations = [row["destination_prefix"] for row in config_index]
+    for destination, count in Counter(destinations).items():
+        if count != 1:
+            failures.append({"check": "unique_destination", "detail": destination})
+    unit_ids = [row["unit_id"] for row in config_index]
+    for unit_id, count in Counter(unit_ids).items():
+        if count != 1:
+            failures.append({"check": "unique_unit_id", "detail": unit_id})
+        if not re.fullmatch(rf"[0-9]{{{EXECUTION_UNIT_INDEX_WIDTH}}}", unit_id):
+            failures.append({"check": "numeric_unit_id", "detail": unit_id})
+    expected_unit_ids = [f"{index:0{EXECUTION_UNIT_INDEX_WIDTH}d}" for index in range(len(config_index))]
+    if unit_ids != expected_unit_ids:
+        failures.append(
+            {
+                "check": "sequential_unit_ids",
+                "detail": f"expected {len(expected_unit_ids):,} globally sequential IDs",
+            }
+        )
+    for row in config_index:
+        destination_index = row.get("destination_index", "")
+        if not re.fullmatch(rf"[0-9]{{{EXECUTION_UNIT_INDEX_WIDTH}}}", destination_index):
+            failures.append(
+                {
+                    "check": "destination_index",
+                    "detail": row["unit_id"],
+                }
+            )
+        expected_destination_suffix = f"/{row['source_layout_prefix']}/{destination_index}"
+        if not row["destination_prefix"].endswith(expected_destination_suffix):
+            failures.append(
+                {
+                    "check": "category_output_destination",
+                    "detail": row["destination_prefix"],
+                }
+            )
+        config_path = build / row["config_path"]
+        with config_path.open() as f:
+            config = yaml.safe_load(f)
+        if config.get("destination_prefix") != row["destination_prefix"]:
+            failures.append(
+                {
+                    "check": "config_destination_matches_index",
+                    "detail": str(config_path),
+                }
+            )
+        if config.get("allow_existing_destination") is not False:
+            failures.append({"check": "no_existing_destination", "detail": str(config_path)})
+        if (
+            int(config.get("max_num_files", 0)) != int(row.get("planned_output_shard_count", 0))
+            or int(row.get("max_num_files", 0)) != int(row.get("planned_output_shard_count", 0))
+            or int(row.get("target_output_shard_bytes", 0)) != target_output_shard_bytes
+            or int(row.get("max_output_shards_per_unit", 0)) != max_output_shards_per_unit
+        ):
+            failures.append(
+                {
+                    "check": "output_shard_policy_matches_config",
+                    "detail": row["unit_id"],
+                }
+            )
+        worker = workers_by_type.get(str(row.get("worker_instance_type", "")))
+        if worker is None:
+            failures.append(
+                {
+                    "check": "worker_instance_in_grid",
+                    "detail": row["unit_id"],
+                }
+            )
+        else:
+            expected_layout = "single" if int(worker["local_nvme_devices"]) == 1 else "raid0"
+            if (
+                int(row.get("worker_vcpus", 0)) != int(worker["vcpus"])
+                or int(row.get("worker_local_nvme_devices", 0)) != int(worker["local_nvme_devices"])
+                or int(row.get("worker_local_nvme_bytes", 0)) != int(worker["local_nvme_bytes"])
+                or row.get("worker_storage_layout") != expected_layout
+            ):
+                failures.append(
+                    {
+                        "check": "worker_resources_match_grid",
+                        "detail": row["unit_id"],
+                    }
+                )
+            if int(row["worker_required_local_nvme_bytes"]) > int(worker["local_nvme_bytes"]):
+                failures.append(
+                    {
+                        "check": "worker_disk_capacity",
+                        "detail": row["unit_id"],
+                    }
+                )
+            work_limit = worker["max_estimated_work_uint32_values"]
+            if work_limit is not None and int(row["estimated_work_uint32_values"]) > int(work_limit):
+                failures.append(
+                    {
+                        "check": "worker_workload_capacity",
+                        "detail": row["unit_id"],
+                    }
+                )
+            if int(config.get("max_workers", 0)) != int(row["max_workers"]) or int(row["max_workers"]) > int(
+                worker["vcpus"]
+            ):
+                failures.append(
+                    {
+                        "check": "worker_concurrency",
+                        "detail": row["unit_id"],
+                    }
+                )
+        manifest_path = config_path.parent / config["source_manifests"][0]["manifest"]
+        if not manifest_path.resolve().is_file():
+            failures.append({"check": "manifest_exists", "detail": str(manifest_path)})
+        else:
+            manifest_rows = _read_csv(manifest_path.resolve())
+            input_view_count = sum(
+                int(manifest_row["repeat_count"])
+                + (int(manifest_row.get("partial_target_uint32_values", 0)) > 0)
+                for manifest_row in manifest_rows
+            )
+            expected_output_shards = _planned_output_shards(
+                output_npy_bytes=int(row["output_npy_bytes"]),
+                input_view_count=input_view_count,
+                target_output_shard_bytes=target_output_shard_bytes,
+                max_output_shards_per_unit=max_output_shards_per_unit,
+            )
+            if (
+                int(row.get("input_view_count", 0)) != input_view_count
+                or int(row.get("planned_output_shard_count", 0)) != expected_output_shards
+                or int(row.get("average_output_shard_bytes", 0))
+                != math.ceil(int(row["output_npy_bytes"]) / expected_output_shards)
+            ):
+                failures.append(
+                    {
+                        "check": "output_shard_count_matches_manifest",
+                        "detail": row["unit_id"],
+                    }
+                )
+            manifest_source_directories = sorted(
+                {manifest_row["npy_uri"].rsplit("/", 1)[0] for manifest_row in manifest_rows}
+            )
+            indexed_source_directories = row["source_directories"].split(";")
+            if manifest_source_directories != indexed_source_directories or len(
+                manifest_source_directories
+            ) != int(row["source_directory_count"]):
+                failures.append(
+                    {
+                        "check": "unit_source_directories_match_manifest",
+                        "detail": str(manifest_path),
+                    }
+                )
+            for manifest_row in manifest_rows:
+                partial_target = int(manifest_row.get("partial_target_uint32_values", 0))
+                if partial_target and (
+                    manifest_row.get("selection_algorithm") != DOCUMENT_SELECTION_ALGORITHM
+                    or not manifest_row.get("selection_seed")
+                ):
+                    failures.append(
+                        {
+                            "check": "partial_selection_manifest",
+                            "detail": str(manifest_path),
+                        }
+                    )
+        launcher_path = build / row["launcher_path"]
+        if not launcher_path.is_file() or not os.access(launcher_path, os.X_OK):
+            failures.append({"check": "launcher_is_executable", "detail": str(launcher_path)})
+        if int(row["estimated_peak_local_bytes"]) > int(row["max_unit_working_bytes"]):
+            failures.append(
+                {
+                    "check": "unit_within_working_budget",
+                    "detail": row["unit_id"],
+                }
+            )
+    active_leaf_ids = {
+        row["leaf_id"]
+        for row in _read_csv(build / "01-plan/resolution/normalized-mix.csv")
+        if row["active"] == "true"
+    }
+    allocated_leaf_ids = {row["leaf_id"] for row in allocations}
+    for leaf_id in sorted(active_leaf_ids - allocated_leaf_ids):
+        failures.append({"check": "active_leaf_allocated", "detail": leaf_id})
+    units_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in config_index:
+        units_by_leaf[row["leaf_id"]].append(row)
+    allocation_by_leaf = {row["leaf_id"]: row for row in allocations}
+    uses_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in object_uses:
+        uses_by_leaf[row["leaf_id"]].append(row)
+    for leaf_id in sorted(active_leaf_ids):
+        units = units_by_leaf.get(leaf_id, [])
+        if not units:
+            failures.append({"check": "category_has_execution_unit", "detail": leaf_id})
+            continue
+        planned = sum(int(row["planned_uint32_values"]) for row in units)
+        target = sum(int(row["target_uint32_values"]) for row in units)
+        allocation = allocation_by_leaf.get(leaf_id)
+        if allocation is None:
+            continue
+        if planned != int(allocation["planned_uint32_values"]):
+            failures.append({"check": "unit_planned_sum", "detail": f"{leaf_id}: {planned}"})
+        if target != int(allocation["target_uint32_values"]):
+            failures.append({"check": "unit_target_sum", "detail": f"{leaf_id}: {target}"})
+        if int(allocation["target_residual_uint32_values"]) != 0:
+            failures.append(
+                {
+                    "check": "exact_proposal_target",
+                    "detail": (f"{leaf_id}: {allocation['target_residual_uint32_values']}"),
+                }
+            )
+        object_planned = sum(int(row["planned_uint32_values"]) for row in uses_by_leaf.get(leaf_id, []))
+        if object_planned != planned:
+            failures.append(
+                {
+                    "check": "object_planned_sum",
+                    "detail": f"{leaf_id}: {object_planned} != {planned}",
+                }
+            )
+    for row in object_uses:
+        source = int(row["estimated_uint32_values"])
+        repeat_count = int(row["repeat_count"])
+        partial_target = int(row.get("partial_target_uint32_values", 0))
+        planned = int(row["planned_uint32_values"])
+        if not 0 <= partial_target < source:
+            failures.append(
+                {
+                    "check": "valid_partial_target",
+                    "detail": f"{row['npy_uri']}: {partial_target} of {source}",
+                }
+            )
+        if planned != source * repeat_count + partial_target:
+            failures.append(
+                {
+                    "check": "object_sampling_arithmetic",
+                    "detail": f"{row['npy_uri']}: {planned}",
+                }
+            )
+    planned_uses = {(row["leaf_id"], row["npy_uri"]) for row in object_uses}
+    required_uses = {
+        (row["leaf_id"], row["npy_uri"])
+        for row in _read_csv(build / "01-plan/inventory/required-objects.csv")
+        if row["active"] == "true"
+    }
+    for leaf_id, uri in sorted(required_uses - planned_uses):
+        failures.append({"check": "required_object_considered", "detail": f"{leaf_id}: {uri}"})
+    _write_csv(phase / "validation-failures.csv", failures, ["check", "detail"])
+    _write_json(
+        phase / "validation-summary.json",
+        {
+            "passed": not failures,
+            "failure_count": len(failures),
+            "config_count": len(config_index),
+            "execution_unit_count": len(config_index),
+            "active_leaf_count": len(active_leaf_ids),
+            "planned_object_rows": len(object_uses),
+        },
+    )
+    if failures:
+        raise PreparationError(f"Proposal validation failed; inspect {phase / 'validation-failures.csv'}")
+
+
+def _filter_execution_units(
+    rows: Sequence[dict[str, str]],
+    *,
+    category: str | None = None,
+    unit: str | None = None,
+) -> list[dict[str, str]]:
+    if category is not None and unit is not None:
+        raise PreparationError(
+            "Choose either a category or an execution unit, not both"
+        )
+    if unit is not None:
+        selected = [row for row in rows if row["unit_id"] == unit]
+        if not selected:
+            raise PreparationError(f"Unknown execution-unit ID: {unit}")
+        return selected
+    if category is None:
+        return list(rows)
+    selected = [row for row in rows if row["leaf_id"] == category]
+    if not selected:
+        selected = [row for row in rows if row["mix_name"] == category]
+    if not selected:
+        selected = [
+            row
+            for row in rows
+            if f"{row['mix_name']}::{row['category_name']}" == category
+        ]
+    if not selected:
+        raise PreparationError(
+            f"Unknown category selector: {category}. "
+            "Use materialize.py --list-categories FILTER to find the exact selector."
+        )
+    return selected
+
+
+def _unit_selection_digest(rows: Sequence[dict[str, str]]) -> str:
+    return hashlib.sha256(
+        "\n".join(sorted(row["unit_id"] for row in rows)).encode()
+    ).hexdigest()
+
+
+def preflight_build(args: argparse.Namespace) -> None:
+    """Re-list the selected inputs and check that their destinations are empty.
+
+    Writes the 02-preflight phase: input-drift.csv, destination-status.csv,
+    preflight-errors.csv, and preflight-summary.json.
+    """
+
+    build = args.build.resolve()
+    manifest = _load_build(build)
+    region = normalize_region(args.region)
+    proposal_summary = build / "01-plan/execution/proposal-summary.json"
+    if not proposal_summary.is_file():
+        raise PreparationError("Proposal is missing; run propose first")
+    _validate_execution_layout(build)
+    phase = _reset_preparation_phase(build, "02-preflight", "03-output-validation")
+    all_config_index = _read_csv(build / "01-plan/execution/config-index.csv")
+    category = getattr(args, "category", None)
+    unit = getattr(args, "unit", None)
+    selected_unit_ids = getattr(args, "selected_unit_ids", None)
+    if selected_unit_ids is not None:
+        requested_ids = set(selected_unit_ids)
+        config_index = [
+            row for row in all_config_index if row["unit_id"] in requested_ids
+        ]
+        resolved_ids = {row["unit_id"] for row in config_index}
+        if not requested_ids:
+            raise PreparationError("Preflight execution-unit selection is empty")
+        if resolved_ids != requested_ids:
+            unknown = sorted(requested_ids - resolved_ids)
+            raise PreparationError(
+                "Preflight received unknown execution-unit IDs: " + ", ".join(unknown)
+            )
+        selection_scope = "explicit_units"
+        selection_value = f"{len(config_index)} execution units"
+    else:
+        config_index = _filter_execution_units(
+            all_config_index, category=category, unit=unit
+        )
+        selection_scope = (
+            "unit" if unit is not None else "category" if category is not None else "all"
+        )
+        selection_value = (
+            unit if unit is not None else category if category is not None else "all"
+        )
+
+    full_listing_plan = _read_csv(build / "01-plan/resolution/listing-plan.csv")
+    approved_inventory = _read_csv(
+        build / "01-plan/inventory/normalized-s3-inventory.csv"
+    )
+    approved_all = {
+        (row["bucket"], row["key"]): row
+        for row in approved_inventory
+        if row["required"] == "true"
+    }
+    if selection_scope == "all":
+        approved = approved_all
+    else:
+        selected_identities: set[tuple[str, str]] = set()
+        manifest_root = (build / "01-plan/execution/manifests").resolve()
+        for config_row in config_index:
+            manifest_path = (build / config_row["manifest_path"]).resolve()
+            if (
+                manifest_path.parent != manifest_root
+                or manifest_path.is_symlink()
+                or not manifest_path.is_file()
+            ):
+                raise PreparationError(
+                    f"Unsafe or missing execution manifest: {manifest_path}"
+                )
+            for manifest_row in _read_csv(manifest_path):
+                for field in ("npy_uri", "metadata_uri"):
+                    parsed = urlparse(manifest_row[field])
+                    if parsed.scheme != "s3" or not parsed.netloc or not parsed.path:
+                        raise PreparationError(
+                            f"Unsupported source URI in {manifest_path}: {manifest_row[field]}"
+                        )
+                    selected_identities.add((parsed.netloc, parsed.path.lstrip("/")))
+        missing_approved = selected_identities - set(approved_all)
+        if missing_approved:
+            raise PreparationError(
+                f"Selected manifests contain {len(missing_approved):,} objects outside the approved inventory"
+            )
+        approved = {
+            identity: approved_all[identity] for identity in selected_identities
+        }
+
+    listing_plan = [
+        row
+        for row in full_listing_plan
+        if any(
+            bucket == row["bucket"] and key.startswith(row["listing_prefix"])
+            for bucket, key in approved
+        )
+    ]
+
+    client, max_workers = _s3_reader(args, manifest, region)
+    current: dict[tuple[str, str], S3Object] = {}
+    errors: list[dict[str, str]] = []
+    with ThreadPoolExecutor(max_workers=max_workers) as pool:
+        futures = {
+            pool.submit(_list_prefix, client, row["bucket"], row["listing_prefix"]): row
+            for row in listing_plan
+        }
+        for future in as_completed(futures):
+            row = futures[future]
+            try:
+                for obj in future.result():
+                    current[(obj.bucket, obj.key)] = obj
+            except Exception as exc:
+                errors.append(
+                    {
+                        "operation": "list",
+                        "bucket": row["bucket"],
+                        "key": row["listing_prefix"],
+                        "error": repr(exc),
+                    }
+                )
+
+    missing = sorted(set(approved) - set(current))
+    if missing:
+        _head_missing_objects(
+            client, missing, current, errors, max_workers=max_workers
+        )
+
+    drift_rows: list[dict[str, Any]] = []
+    for identity, expected in sorted(approved.items()):
+        actual = current.get(identity)
+        if actual is None:
+            status = "missing"
+        elif actual.size_bytes != int(expected["size_bytes"]):
+            status = "size_changed"
+        elif expected["etag"] and actual.etag != expected["etag"]:
+            status = "etag_changed"
+        elif expected["last_modified"] and _normalize_timestamp(
+            actual.last_modified
+        ) != _normalize_timestamp(expected["last_modified"]):
+            status = "last_modified_changed"
+        else:
+            status = "unchanged"
+        drift_rows.append(
+            {
+                "bucket": identity[0],
+                "key": identity[1],
+                "status": status,
+                "expected_size_bytes": expected["size_bytes"],
+                "actual_size_bytes": actual.size_bytes if actual else "",
+                "expected_etag": expected["etag"],
+                "actual_etag": actual.etag if actual else "",
+                "expected_last_modified": expected["last_modified"],
+                "actual_last_modified": actual.last_modified if actual else "",
+            }
+        )
+
+    destination_rows: list[dict[str, str]] = []
+    with ThreadPoolExecutor(max_workers=max_workers) as pool:
+        futures = {}
+        for row in config_index:
+            parsed = urlparse(row["destination_prefix"])
+            prefix = parsed.path.lstrip("/").rstrip("/") + "/"
+            future = pool.submit(
+                client.list_objects_v2,
+                Bucket=parsed.netloc,
+                Prefix=prefix,
+                MaxKeys=1,
+            )
+            futures[future] = (row, parsed.netloc, prefix)
+        for future in as_completed(futures):
+            row, bucket, prefix = futures[future]
+            try:
+                response = future.result()
+                contents = response.get("Contents", [])
+                destination_rows.append(
+                    {
+                        "unit_id": row["unit_id"],
+                        "leaf_id": row["leaf_id"],
+                        "mix_name": row["mix_name"],
+                        "category_name": row["category_name"],
+                        "source_directories": row["source_directories"],
+                        "destination_prefix": row["destination_prefix"],
+                        "status": "occupied" if contents else "empty",
+                        "first_existing_key": contents[0]["Key"] if contents else "",
+                    }
+                )
+            except Exception as exc:
+                errors.append(
+                    {
+                        "operation": "destination_list",
+                        "bucket": bucket,
+                        "key": prefix,
+                        "error": repr(exc),
+                    }
+                )
+
+    drift_fields = [
+        "bucket",
+        "key",
+        "status",
+        "expected_size_bytes",
+        "actual_size_bytes",
+        "expected_etag",
+        "actual_etag",
+        "expected_last_modified",
+        "actual_last_modified",
+    ]
+    _write_csv(phase / "input-drift.csv", drift_rows, drift_fields)
+    _write_csv(
+        phase / "destination-status.csv",
+        sorted(destination_rows, key=lambda row: row["destination_prefix"]),
+        [
+            "unit_id",
+            "leaf_id",
+            "mix_name",
+            "category_name",
+            "source_directories",
+            "destination_prefix",
+            "status",
+            "first_existing_key",
+        ],
+    )
+    _write_csv(
+        phase / "preflight-errors.csv", errors, ["operation", "bucket", "key", "error"]
+    )
+    drifted = sum(row["status"] != "unchanged" for row in drift_rows)
+    occupied = sum(row["status"] != "empty" for row in destination_rows)
+    passed = (
+        not errors
+        and not drifted
+        and not occupied
+        and len(destination_rows) == len(config_index)
+    )
+    summary = {
+        "created_at": _utc_now(),
+        "selection_scope": selection_scope,
+        "selection_value": selection_value,
+        "selected_execution_units": len(config_index),
+        "total_execution_units": len(all_config_index),
+        "selected_unit_ids_sha256": _unit_selection_digest(config_index),
+        "approved_input_objects": len(approved),
+        "drifted_input_objects": drifted,
+        "destinations_checked": len(destination_rows),
+        "occupied_destinations": occupied,
+        "errors": len(errors),
+        "passed": passed,
+        "materialization_executed": False,
+    }
+    _write_json(phase / "preflight-summary.json", summary)
+    if not passed:
+        raise PreparationError(f"Preflight failed; inspect artifacts in {phase}")
+    if not getattr(args, "quiet", False):
+        print(
+            f"Preflight passed for {len(config_index):,} selected execution unit(s): {phase}"
+        )
+
+
+def verify_output(args: argparse.Namespace) -> None:
+    """Verify materialized outputs from S3 object listings and sizes alone.
+
+    Writes the 03-output-validation phase: output-inventory.csv,
+    output-validation.csv, output-problems.csv, output-errors.csv,
+    output-summary.json, plot-data/source-target-actual.csv, and report.html.
+    """
+
+    build = args.build.resolve()
+    manifest = _load_build(build)
+    region = normalize_region(args.region)
+    _validate_execution_layout(build)
+    settings = manifest["settings"]
+    config_index = _read_csv(build / "01-plan/execution/config-index.csv")
+    if not config_index:
+        raise PreparationError("Proposal config index is empty or missing")
+    category = getattr(args, "category", None)
+    unit = getattr(args, "unit", None)
+    selection_scope = "all"
+    if category:
+        config_index = _filter_execution_units(config_index, category=category)
+        selection_scope = f"category:{category}"
+    elif unit:
+        config_index = _filter_execution_units(config_index, unit=unit)
+        selection_scope = f"unit:{unit}"
+    phase = _reset_preparation_phase(build, "03-output-validation")
+    client, max_workers = _s3_reader(args, manifest, region)
+    output_objects: list[dict[str, Any]] = []
+    validation_rows: list[dict[str, Any]] = []
+    unexpected_rows: list[dict[str, str]] = []
+    errors: list[dict[str, str]] = []
+
+    with ThreadPoolExecutor(max_workers=max_workers) as pool:
+        futures = {}
+        for row in config_index:
+            parsed = urlparse(row["destination_prefix"])
+            prefix = parsed.path.lstrip("/").rstrip("/") + "/"
+            future = pool.submit(_list_prefix, client, parsed.netloc, prefix)
+            futures[future] = (row, parsed.netloc, prefix)
+        for future in as_completed(futures):
+            row, bucket, prefix = futures[future]
+            try:
+                objects = future.result()
+            except Exception as exc:
+                errors.append(
+                    {
+                        "operation": "list_output",
+                        "bucket": bucket,
+                        "key": prefix,
+                        "error": repr(exc),
+                    }
+                )
+                continue
+            object_map = {obj.key: obj for obj in objects}
+            npys = [obj for obj in objects if obj.key.endswith(".npy")]
+            metadata = {obj.key for obj in objects if obj.key.endswith(".csv.gz")}
+            unexpected = [
+                obj for obj in objects if not obj.key.endswith((".npy", ".csv.gz"))
+            ]
+            missing_metadata = [
+                _pair_metadata_key(obj.key)
+                for obj in npys
+                if _pair_metadata_key(obj.key) not in metadata
+            ]
+            orphan_metadata = [
+                key
+                for key in metadata
+                if key[: -len(".csv.gz")] + ".npy" not in object_map
+            ]
+            invalid_npys = [
+                obj
+                for obj in npys
+                if obj.size_bytes <= 0 or obj.size_bytes % UINT32_BYTES
+            ]
+            actual = sum(obj.size_bytes // UINT32_BYTES for obj in npys)
+            predicted = int(row["planned_uint32_values"])
+            planned_output_shards = int(row["planned_output_shard_count"])
+            allowed_residual = int(
+                row["allowed_materialized_target_residual_uint32_values"]
+            )
+            status = (
+                "passed"
+                if len(npys) == planned_output_shards
+                and not missing_metadata
+                and not orphan_metadata
+                and not invalid_npys
+                and not unexpected
+                and abs(actual - predicted) <= allowed_residual
+                else "failed"
+            )
+            validation_rows.append(
+                {
+                    "unit_id": row["unit_id"],
+                    "leaf_id": row["leaf_id"],
+                    "mix_index": row["mix_index"],
+                    "mix_name": row["mix_name"],
+                    "category_index": row["category_index"],
+                    "category_name": row["category_name"],
+                    "destination_prefix": row["destination_prefix"],
+                    "target_uint32_values": row["target_uint32_values"],
+                    "predicted_uint32_values": predicted,
+                    "actual_uint32_values": actual,
+                    "actual_minus_predicted": actual - predicted,
+                    "allowed_target_residual_uint32_values": allowed_residual,
+                    "planned_output_shard_count": planned_output_shards,
+                    "npy_count": len(npys),
+                    "metadata_count": len(metadata),
+                    "missing_metadata_count": len(missing_metadata),
+                    "orphan_metadata_count": len(orphan_metadata),
+                    "invalid_npy_count": len(invalid_npys),
+                    "unexpected_object_count": len(unexpected),
+                    "status": status,
+                }
+            )
+            for obj in objects:
+                output_objects.append(
+                    {
+                        "unit_id": row["unit_id"],
+                        "mix_name": row["mix_name"],
+                        "category_name": row["category_name"],
+                        "destination_prefix": row["destination_prefix"],
+                        "bucket": obj.bucket,
+                        "key": obj.key,
+                        "size_bytes": obj.size_bytes,
+                        "etag": obj.etag,
+                        "last_modified": obj.last_modified,
+                        "object_type": (
+                            "npy"
+                            if obj.key.endswith(".npy")
+                            else "metadata"
+                            if obj.key.endswith(".csv.gz")
+                            else "unexpected"
+                        ),
+                    }
+                )
+            for key in missing_metadata:
+                unexpected_rows.append(
+                    {
+                        "mix_name": row["mix_name"],
+                        "key": key,
+                        "reason": "missing metadata partner",
+                    }
+                )
+            for key in orphan_metadata:
+                unexpected_rows.append(
+                    {
+                        "mix_name": row["mix_name"],
+                        "key": key,
+                        "reason": "orphan metadata",
+                    }
+                )
+            for obj in invalid_npys:
+                unexpected_rows.append(
+                    {
+                        "mix_name": row["mix_name"],
+                        "key": obj.key,
+                        "reason": "invalid NPY size",
+                    }
+                )
+            for obj in unexpected:
+                unexpected_rows.append(
+                    {
+                        "mix_name": row["mix_name"],
+                        "key": obj.key,
+                        "reason": "unexpected output object type",
+                    }
+                )
+
+    validation_rows.sort(key=lambda row: row["unit_id"])
+    _write_csv(
+        phase / "output-inventory.csv",
+        output_objects,
+        [
+            "unit_id",
+            "mix_name",
+            "category_name",
+            "destination_prefix",
+            "bucket",
+            "key",
+            "size_bytes",
+            "etag",
+            "last_modified",
+            "object_type",
+        ],
+    )
+    validation_fields = [
+        "unit_id",
+        "leaf_id",
+        "mix_index",
+        "mix_name",
+        "category_index",
+        "category_name",
+        "destination_prefix",
+        "target_uint32_values",
+        "predicted_uint32_values",
+        "actual_uint32_values",
+        "actual_minus_predicted",
+        "allowed_target_residual_uint32_values",
+        "planned_output_shard_count",
+        "npy_count",
+        "metadata_count",
+        "missing_metadata_count",
+        "orphan_metadata_count",
+        "invalid_npy_count",
+        "unexpected_object_count",
+        "status",
+    ]
+    _write_csv(phase / "output-validation.csv", validation_rows, validation_fields)
+    _write_csv(
+        phase / "output-problems.csv",
+        unexpected_rows,
+        ["mix_name", "key", "reason"],
+    )
+    _write_csv(
+        phase / "output-errors.csv", errors, ["operation", "bucket", "key", "error"]
+    )
+    failed = sum(row["status"] != "passed" for row in validation_rows)
+    target_total = sum(int(row["target_uint32_values"]) for row in validation_rows)
+    actual_total = sum(int(row["actual_uint32_values"]) for row in validation_rows)
+    aggregate_residual = actual_total - target_total
+    allowed_aggregate_residual = math.ceil(
+        target_total
+        * float(settings["max_materialized_total_target_residual_fraction"])
+    )
+    aggregate_within_bound = abs(aggregate_residual) <= allowed_aggregate_residual
+    passed = (
+        not errors
+        and not failed
+        and len(validation_rows) == len(config_index)
+        and aggregate_within_bound
+    )
+    _write_json(
+        phase / "output-summary.json",
+        {
+            "created_at": _utc_now(),
+            "selection_scope": selection_scope,
+            "selected_unit_ids_sha256": _unit_selection_digest(config_index),
+            "destinations_expected": len(config_index),
+            "destinations_checked": len(validation_rows),
+            "failed_destinations": failed,
+            "errors": len(errors),
+            "target_uint32_values": target_total,
+            "predicted_uint32_values": sum(
+                int(row["predicted_uint32_values"]) for row in validation_rows
+            ),
+            "actual_uint32_values": actual_total,
+            "planned_output_shard_count": sum(
+                int(row["planned_output_shard_count"]) for row in validation_rows
+            ),
+            "actual_output_shard_count": sum(int(row["npy_count"]) for row in validation_rows),
+            "actual_output_file_count": sum(
+                int(row["npy_count"]) + int(row["metadata_count"]) for row in validation_rows
+            ),
+            "materialized_target_residual_uint32_values": aggregate_residual,
+            "allowed_materialized_target_residual_uint32_values": (
+                allowed_aggregate_residual
+            ),
+            "aggregate_target_residual_within_bound": aggregate_within_bound,
+            "passed": passed,
+        },
+    )
+    plot_data = phase / "plot-data"
+    plot_data.mkdir(exist_ok=False)
+    inventory_details_path = build / "01-plan/inventory/inventory-details.json"
+    if inventory_details_path.is_symlink() or not inventory_details_path.is_file():
+        raise PreparationError(
+            f"Inventory hierarchy is missing or unsafe: {inventory_details_path}"
+        )
+    with inventory_details_path.open(encoding="utf-8") as handle:
+        inventory_details = json.load(handle)
+    hierarchy_rows, hierarchy_html = render_output_validation_report(
+        inventory_details=inventory_details,
+        validation_rows=validation_rows,
+    )
+    _write_csv(
+        plot_data / "source-target-actual.csv",
+        hierarchy_rows,
+        [
+            "level",
+            "source_family",
+            "subcategory_name",
+            "category_name",
+            "lower_group",
+            "leaf_id",
+            "source_uint32_values",
+            "target_uint32_values",
+            "actual_uint32_values",
+            "actual_minus_target",
+            "planned_sampling_ratio",
+            "realized_sampling_ratio",
+            "status",
+        ],
+    )
+    _write_text(phase / "report.html", hierarchy_html)
+    if not passed:
+        raise PreparationError(
+            f"Output validation failed; inspect artifacts in {phase}"
+        )
+    print(f"Output validation passed using size-only checks: {phase}")
+
+
+def _svg_bar_chart(
+    title: str,
+    labels: Sequence[str],
+    values: Sequence[float],
+    unit: str,
+    width: int = 1100,
+    value_labels: Sequence[str] | None = None,
+    summary: str | None = None,
+) -> str:
+    if value_labels is not None and len(value_labels) != len(values):
+        raise ValueError("value_labels must have the same length as values")
+    row_height = 24
+    label_x = 12
+    margin_left = 390
+    margin_right = 240 if value_labels is not None else 140
+    height = 70 + row_height * len(labels)
+    plot_width = width - margin_left - margin_right
+    maximum = max(values, default=1.0) or 1.0
+    rows = []
+    for index, (label, value) in enumerate(zip(labels, values)):
+        y = 50 + index * row_height
+        bar_width = max(0.0, plot_width * value / maximum)
+        displayed_value = (
+            value_labels[index] if value_labels is not None else f"{value:.4g} {unit}"
+        )
+        rows.append(
+            f'<text x="{label_x}" y="{y + 14}">{html.escape(label[:52])}</text>'
+            f'<rect x="{margin_left}" y="{y}" width="{bar_width:.2f}" height="16" />'
+            f'<text x="{margin_left + plot_width + 8}" y="{y + 14}">{html.escape(displayed_value)}</text>'
+        )
+    return (
+        f'<svg xmlns="http://www.w3.org/2000/svg" width="{width}" height="{height}" role="img" '
+        f'aria-label="{html.escape(title)}"><title>{html.escape(title)}</title><style>'
+        "text{font:12px sans-serif;fill:#222}rect{fill:#356cb6}"
+        ".summary{font-size:14px;font-weight:600}</style>"
+        + (
+            f'<text class="summary" x="{label_x}" y="24">{html.escape(summary)}</text>'
+            if summary
+            else ""
+        )
+        + "".join(rows)
+        + "</svg>\n"
+    )
+
+
+def _human_token_count(value: int) -> str:
+    for scale, suffix in (
+        (1_000_000_000_000, "T"),
+        (1_000_000_000, "B"),
+        (1_000_000, "M"),
+        (1_000, "K"),
+    ):
+        if abs(value) >= scale:
+            return f"{value / scale:.3g}{suffix}"
+    return f"{value:,}"
+
+
+def _sampling_change(original: int, target: int) -> tuple[str, str, float | None]:
+    if original <= 0:
+        if target <= 0:
+            return "no source tokens", "sampling-same", None
+        return "no source tokens", "sampling-unknown", None
+    ratio = target / original
+    if math.isclose(ratio, 1.0, rel_tol=0.0, abs_tol=0.0005):
+        return "1.00× unchanged", "sampling-same", ratio
+    direction = "upsample" if ratio > 1 else "downsample"
+    css_class = "sampling-up" if ratio > 1 else "sampling-down"
+    return f"{_format_multiplier(ratio)} {direction}", css_class, ratio
+
+
+def _sampling_rate_label(original: int, target: int) -> tuple[str, str]:
+    label, css_class, _ = _sampling_change(original, target)
+    return label, css_class
+
+
+def _format_multiplier(value: float) -> str:
+    if value >= 100:
+        return f"{value:,.0f}×"
+    if value >= 10:
+        return f"{value:.1f}×"
+    return f"{value:.2f}×"
+
+
+def _apportion_by_size(total: int, sizes: Sequence[int]) -> list[int]:
+    denominator = sum(sizes)
+    if total <= 0 or denominator <= 0:
+        return [0 for _ in sizes]
+    apportioned = [total * size // denominator for size in sizes]
+    remainder = total - sum(apportioned)
+    order = sorted(
+        range(len(sizes)),
+        key=lambda index: (-(total * sizes[index] % denominator), index),
+    )
+    for index in order[:remainder]:
+        apportioned[index] += 1
+    return apportioned
+
+
+def _comparison_bars(original: int, target: int) -> str:
+    maximum = max(original, target, 1)
+    original_width = 100 * original / maximum
+    target_width = 100 * target / maximum
+    return (
+        '<span class="comparison-bars" aria-hidden="true">'
+        '<span class="comparison-track"><span class="original-fill" '
+        f'style="width:{original_width:.8f}%"></span></span>'
+        '<span class="comparison-track"><span class="target-fill" '
+        f'style="width:{target_width:.8f}%"></span></span></span>'
+    )
+
+
+def _mix_plot_value_label(value: int, total: int) -> str:
+    percent = 100 * value / total if total else 0.0
+    return f"{percent:.2f}% · {_human_token_count(value)} tokens"
+
+
+def _summary_metrics(metrics: Sequence[tuple[str, str]]) -> str:
+    return (
+        '<div class="summary-metrics">'
+        + "".join(
+            '<div class="summary-metric">'
+            f'<span class="summary-label">{html.escape(label)}</span>'
+            f'<span class="summary-value">{html.escape(value)}</span>'
+            "</div>"
+            for label, value in metrics
+        )
+        + "</div>"
+    )
+
+
+def _path_subgroup(yaml_path: str) -> str:
+    parts = [part for part in yaml_path.split("/") if part]
+    for index, part in enumerate(parts):
+        if part == "allenai" and index:
+            return unquote(parts[index - 1])
+    candidates = [
+        part
+        for part in parts
+        if "*" not in part and not part.endswith((".npy", ".csv.gz"))
+    ]
+    return unquote(candidates[-1]) if candidates else yaml_path
+
+
+def _source_uri_details(source_uris: Iterable[str], empty_message: str) -> str:
+    resolved_sources = sorted(set(source_uris)) or [empty_message]
+    return "".join(
+        f'<span class="path-detail-uri">{html.escape(uri)}</span>'
+        for uri in resolved_sources
+    )
+
+
+def _split_mix_name(mix_name: str) -> tuple[str, str]:
+    source_family, separator, subcategory = mix_name.partition(":")
+    return source_family, subcategory if separator else "default"
+
+
+def _interactive_report_style() -> str:
+    return (
+        "\n<style>\n"
+        + report_root_style(
+            (
+                ("muted", "muted"),
+                ("surface-hover", "surface-hover"),
+                ("surface-selected", "surface-selected"),
+                ("detail", "surface"),
+                ("track", "track"),
+                ("series", "accent"),
+                ("original", "neutral-series"),
+                ("up", "accent"),
+                ("down", "warning"),
+                ("same", "muted"),
+                ("code", "code"),
+            )
+        )
+        + """
+*{box-sizing:border-box}body{font:14px/1.45 system-ui,sans-serif;max-width:1120px;margin:0 auto;padding:32px 24px 72px;color:CanvasText;background:Canvas}h1{margin:0;font-size:26px;line-height:1.2}h2{margin:0;font-size:20px;line-height:1.3;overflow-wrap:anywhere}.chart-total{margin:8px 0 24px;color:var(--muted);font-variant-numeric:tabular-nums}.summary-metrics{display:grid;grid-template-columns:repeat(auto-fit,minmax(150px,1fr));gap:12px 32px;margin:18px 10px 26px;font-variant-numeric:tabular-nums}.summary-label,.mix-metric-label{display:block;margin-bottom:2px;color:var(--muted)}.summary-value{display:block;font-size:18px;font-weight:500}.mix-chart{display:grid;gap:2px}
+.mix-row,.subcategory-row{display:grid;grid-template-columns:minmax(0,1fr) auto;gap:8px 20px;width:100%;padding:12px 10px;border:0;border-radius:8px;background:transparent;color:inherit;text-align:left;font:inherit;cursor:pointer}.mix-row:hover,.subcategory-row:hover{background:var(--surface-hover)}.mix-row.is-selected,.subcategory-row.is-selected{background:var(--surface-selected)}.mix-name{min-width:0;overflow-wrap:anywhere;font-weight:500}.mix-value{white-space:nowrap;color:var(--muted);font-variant-numeric:tabular-nums}.bar-track{grid-column:1/-1;display:block;height:6px;overflow:hidden;background:var(--track);border-radius:999px}.bar-fill{display:block;height:100%;background:var(--series);border-radius:inherit}
+.mix-row.has-metrics,.subcategory-row.has-metrics{grid-template-columns:1fr;gap:8px}.mix-metrics{display:grid;grid-template-columns:repeat(auto-fit,minmax(140px,1fr));gap:5px 24px;color:var(--muted);font-variant-numeric:tabular-nums}.mix-metric-value{color:CanvasText}.subcategory-list{display:grid;gap:2px}.subcategory-detail{padding:20px 10px 26px}.subcategory-detail .detail-head{margin-bottom:18px}
+.mix-detail{padding:22px 18px 28px;border-radius:10px;background:var(--detail)}.mix-detail[hidden],.subcategory-detail[hidden]{display:none}.detail-head{display:grid;grid-template-columns:minmax(0,1fr) auto;gap:8px 20px;align-items:end;margin-bottom:20px}.detail-total{color:var(--muted);font-variant-numeric:tabular-nums;text-align:right}.category-grid{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:24px 32px}.category{min-width:0}.category-head{display:grid;grid-template-columns:minmax(0,1fr) auto;gap:12px;align-items:baseline}.category-name{font-weight:500;overflow-wrap:anywhere}.sampling{font-variant-numeric:tabular-nums;white-space:nowrap}.sampling-up{color:var(--up)}.sampling-down{color:var(--down)}.sampling-same,.sampling-unknown{color:var(--same)}.category-metrics{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:5px 24px;margin:4px 0 7px;color:var(--muted);font-variant-numeric:tabular-nums}.category-metrics.proposal-metrics{grid-template-columns:repeat(2,minmax(0,1fr))}.category-metric{min-width:0}.category-bar{height:4px;overflow:hidden;margin-top:7px;background:var(--track);border-radius:999px}.category-bar-fill{display:block;height:100%;background:var(--series);border-radius:inherit}.comparison-bars{display:grid;grid-template-rows:3px 3px;gap:3px}.comparison-track{display:block;overflow:hidden;background:var(--track);border-radius:999px}.original-fill,.target-fill{display:block;height:100%;border-radius:inherit}.original-fill{background:var(--original)}.target-fill{background:var(--series)}.path-list{display:grid;gap:2px;margin-top:8px}.path-detail summary{display:grid;grid-template-columns:minmax(0,1fr) auto;gap:5px 12px;padding:8px 6px;border-radius:6px;cursor:pointer;list-style-position:inside}.path-detail summary:hover{background:var(--surface-hover)}.path-name{overflow-wrap:anywhere}.path-stat{color:var(--muted);text-align:right;white-space:nowrap;font-variant-numeric:tabular-nums}.path-sampling{grid-column:1/-1;display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:5px 24px;color:var(--muted);font-variant-numeric:tabular-nums}.path-metric{min-width:0}.path-sampling .comparison-bars,.path-use-summary{grid-column:1/-1}.path-chevron{display:inline-block;margin-left:7px;transition:transform .12s ease}.path-detail[open] .path-chevron{transform:rotate(90deg)}.path-detail code{display:block;margin:2px 6px 10px;padding:11px 12px;border-radius:6px;background:var(--code);font:12px/1.45 ui-monospace,monospace;overflow-wrap:anywhere}.path-detail-metrics{display:flex;flex-wrap:wrap;gap:3px 24px}.path-detail-metric{white-space:nowrap}.path-detail-uri{display:block;margin-top:7px}.audit-warning{margin:8px 0 24px;padding:16px 18px;border-left:4px solid var(--down);border-radius:6px;background:var(--detail)}.audit-warning h2{font-size:17px}.audit-warning p{margin:6px 0 10px}.audit-warning ul{margin:0;padding-left:20px;font-variant-numeric:tabular-nums}.supporting-plots{display:grid;grid-template-columns:1fr 1fr;gap:24px;margin-top:32px}.supporting-plots img{display:block;width:100%;height:auto}
+.repetition-line{display:flex;flex-wrap:wrap;gap:5px 12px;margin-top:7px;color:var(--muted);font-variant-numeric:tabular-nums}
+@media(prefers-reduced-motion:reduce){.path-chevron{transition:none}}
+@media(max-width:760px){body{padding:24px 16px 48px}.summary-metrics{margin-left:6px;margin-right:6px}.mix-row,.subcategory-row{grid-template-columns:1fr;gap:7px}.mix-value{white-space:normal}.bar-track{grid-column:1}.detail-head{grid-template-columns:1fr}.detail-total{white-space:normal;text-align:left}.category-grid{grid-template-columns:1fr}.category-metrics,.category-metrics.proposal-metrics,.path-sampling{grid-template-columns:1fr}.supporting-plots{grid-template-columns:1fr}}
+</style>
+"""
+    )
+
+
+def _interactive_chart_rows(
+    rows: Sequence[dict[str, Any]],
+    value_field: str,
+    percent_field: str,
+    detail_prefix: str,
+    total: int,
+    row_class: str = "mix-row",
+) -> str:
+    maximum = max((float(row[percent_field]) for row in rows), default=1.0) or 1.0
+    output: list[str] = []
+    for index, row in enumerate(rows):
+        detail_id = f"{detail_prefix}-{index}"
+        row["detail_id"] = detail_id
+        percent = float(row[percent_field])
+        relative_width = 100 * percent / maximum
+        tokens = int(row[value_field])
+        value_label = str(
+            row.get("value_label") or _mix_plot_value_label(tokens, total)
+        )
+        metric_columns = row.get("metric_columns")
+        if metric_columns:
+            metrics = (
+                '<span class="mix-metrics">'
+                + "".join(
+                    '<span class="mix-metric">'
+                    f'<span class="mix-metric-label">{html.escape(str(metric["label"]))}</span>'
+                    f'<span class="mix-metric-value">{html.escape(str(metric["value"]))}</span>'
+                    "</span>"
+                    for metric in metric_columns
+                )
+                + "</span>"
+            )
+            rendered_row_class = f"{row_class} has-metrics"
+            value = metrics
+        else:
+            rendered_row_class = row_class
+            value = f'<span class="mix-value">{html.escape(value_label)}</span>'
+        output.append(
+            f'<button type="button" class="{rendered_row_class}" data-detail="{detail_id}" '
+            f'aria-controls="{detail_id}" aria-expanded="false">'
+            f'<span class="mix-name">{html.escape(str(row.get("display_name", row["mix_name"])))}</span>'
+            + value
+            + f'<span class="bar-track" aria-hidden="true"><span class="bar-fill" style="width:{relative_width:.8f}%"></span></span>'
+            "</button>"
+        )
+    return "".join(output)
+
+
+def _interactive_report_script() -> str:
+    return """
+<script>
+const setAccordionState = (button, detail, expanded) => {
+  button.classList.toggle('is-selected', expanded);
+  button.setAttribute('aria-expanded', String(expanded));
+  detail.hidden = !expanded;
+};
+document.querySelectorAll('.mix-row').forEach((button) => {
+  button.addEventListener('click', () => {
+    const target = document.getElementById(button.dataset.detail);
+    const opening = button.getAttribute('aria-expanded') !== 'true';
+    document.querySelectorAll('.mix-row').forEach((row) => {
+      const detail = document.getElementById(row.dataset.detail);
+      if (detail) setAccordionState(row, detail, false);
+    });
+    document.querySelectorAll('.subcategory-row').forEach((row) => {
+      const detail = document.getElementById(row.dataset.detail);
+      if (detail) setAccordionState(row, detail, false);
+    });
+    if (opening) {
+      button.insertAdjacentElement('afterend', target);
+      setAccordionState(button, target, true);
+      target.scrollIntoView({block: 'nearest'});
+    }
+  });
+});
+document.querySelectorAll('.subcategory-row').forEach((button) => {
+  button.addEventListener('click', () => {
+    const family = button.closest('.mix-detail');
+    const target = document.getElementById(button.dataset.detail);
+    const opening = button.getAttribute('aria-expanded') !== 'true';
+    family.querySelectorAll('.subcategory-row').forEach((row) => {
+      const detail = document.getElementById(row.dataset.detail);
+      if (detail) setAccordionState(row, detail, false);
+    });
+    if (opening) {
+      button.insertAdjacentElement('afterend', target);
+      setAccordionState(button, target, true);
+      target.scrollIntoView({block: 'nearest'});
+    }
+  });
+});
+</script>
+"""
+
+
+def _human_byte_count(value: int) -> str:
+    for scale, suffix in (
+        (1_000_000_000_000, "TB"),
+        (1_000_000_000, "GB"),
+        (1_000_000, "MB"),
+        (1_000, "KB"),
+    ):
+        if abs(value) >= scale:
+            return f"{value / scale:.3g} {suffix}"
+    return f"{value:,} B"
+
+
+def _percentile(values: Sequence[int], percentile: float) -> int:
+    if not values:
+        return 0
+    ordered = sorted(values)
+    index = max(0, min(len(ordered) - 1, math.ceil(percentile * len(ordered)) - 1))
+    return ordered[index]
+
+
+def _proposal_artifact_href(path: Any) -> str:
+    relative = Path(str(path)).as_posix()
+    prefix = "01-plan/execution/"
+    if relative.startswith(prefix):
+        relative = relative[len(prefix) :]
+    return html.escape(relative, quote=True)
+
+
+def _execution_proposal_style() -> str:
+    return (
+        "\n<style>\n"
+        + report_root_style(
+            (
+                ("muted", "muted"),
+                ("surface", "surface"),
+                ("surface-hover", "surface-raised"),
+                ("track", "track"),
+                ("source", "neutral-series"),
+                ("output", "accent-output"),
+                ("selection", "warning-selection"),
+                ("link", "accent-link"),
+            )
+        )
+        + """
+*{box-sizing:border-box}body{max-width:1240px;margin:0 auto;padding:34px 26px 72px;background:Canvas;color:CanvasText;font:14px/1.45 system-ui,sans-serif}h1{margin:0;font-size:28px;line-height:1.2}h2{margin:34px 0 14px;font-size:20px}.lede{max-width:820px;margin:8px 0 0;color:var(--muted)}.execution-metrics{display:grid;grid-template-columns:repeat(auto-fit,minmax(150px,1fr));gap:12px 28px;margin:20px 0}.execution-metric{min-width:0}.metric-label{display:block;color:var(--muted)}.metric-value{display:block;margin-top:2px;font-size:18px;font-weight:600;font-variant-numeric:tabular-nums}.storage-note{margin:0 0 16px;color:var(--muted)}.worker-grid{width:100%;border-collapse:separate;border-spacing:0 6px;margin:0}.worker-grid th{padding:0 12px 4px;color:var(--muted);font-size:12px;font-weight:500;text-align:left}.worker-grid td{padding:11px 12px;background:var(--surface);font-variant-numeric:tabular-nums}.worker-grid td:first-child{border-radius:8px 0 0 8px;font-weight:600}.worker-grid td:last-child{border-radius:0 8px 8px 0}.split-list{display:grid;gap:6px}.split-category{display:grid;grid-template-columns:minmax(0,1fr) repeat(4,minmax(115px,auto));gap:12px 24px;align-items:center;padding:11px 14px;border-radius:8px;background:var(--surface)}.split-name{overflow-wrap:anywhere;font-weight:600}.split-value{font-variant-numeric:tabular-nums}.split-value span{display:block;color:var(--muted);font-size:12px}.unit-heading{display:flex;flex-wrap:wrap;gap:10px 20px;align-items:end;justify-content:space-between}.unit-heading h2{margin-bottom:0}.visible-count{color:var(--muted);font-variant-numeric:tabular-nums}.unit-controls{display:grid;grid-template-columns:minmax(220px,1fr) auto;gap:10px 20px;margin:14px 0}.unit-controls input[type=search]{width:100%;padding:9px 11px;border:0;border-radius:7px;background:var(--surface);color:inherit;font:inherit}.unit-controls label{display:flex;gap:8px;align-items:center;color:var(--muted)}.unit-list{display:grid;gap:6px}.execution-unit{border:0;border-radius:9px;background:var(--surface)}.execution-unit[hidden]{display:none}.execution-unit summary{display:grid;grid-template-columns:minmax(0,1fr) minmax(210px,290px);gap:10px 28px;padding:14px 16px;cursor:pointer;list-style-position:inside}.execution-unit summary:hover{background:var(--surface-hover);border-radius:9px}.unit-title{min-width:0;overflow-wrap:anywhere;font-weight:600}.unit-position{display:block;margin:2px 0 0 18px;color:var(--muted);font-size:12px;font-weight:400}.unit-disk{font-variant-numeric:tabular-nums}.unit-disk strong,.unit-disk span{display:block}.unit-disk span{color:var(--muted);font-size:12px}.unit-body{padding:2px 16px 17px}.unit-metrics{display:grid;grid-template-columns:repeat(auto-fit,minmax(140px,1fr));gap:12px 24px;margin:6px 0 16px}.unit-metric{min-width:0}.unit-metric span{display:block;color:var(--muted);font-size:12px}.unit-metric strong{display:block;margin-top:2px;font-weight:600;font-variant-numeric:tabular-nums}.disk-breakdown{display:flex;height:9px;overflow:hidden;border-radius:999px;background:var(--track)}.disk-segment{display:block;height:100%}.disk-source{background:var(--source)}.disk-output{background:var(--output)}.disk-selection{background:var(--selection)}.disk-legend{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:7px 22px;margin:7px 0 16px;color:var(--muted);font-variant-numeric:tabular-nums}.disk-legend span::before{display:inline-block;width:8px;height:8px;margin-right:6px;border-radius:2px;content:""}.legend-source::before{background:var(--source)}.legend-output::before{background:var(--output)}.legend-selection::before{background:var(--selection)}.unit-paths{display:grid;gap:8px;margin:0}.unit-paths div{min-width:0}.unit-paths dt{color:var(--muted);font-size:12px}.unit-paths dd{margin:2px 0 0}.unit-paths code{display:block;padding:8px 10px;border-radius:6px;background:Canvas;overflow-wrap:anywhere;font:12px/1.4 ui-monospace,monospace}.artifact-links{display:flex;flex-wrap:wrap;gap:8px 16px;margin-top:12px}.artifact-links a{color:var(--link);font-weight:600;text-decoration:none}.artifact-links a:hover{text-decoration:underline}
+@media(max-width:760px){body{padding:24px 16px 48px}.worker-grid{display:block;overflow-x:auto}.split-category{grid-template-columns:1fr 1fr}.execution-unit summary{grid-template-columns:1fr}.unit-metrics{grid-template-columns:repeat(2,1fr)}.disk-legend{grid-template-columns:1fr}.unit-controls{grid-template-columns:1fr}}
+</style>
+"""
+    )
+
+
+def _execution_proposal_script() -> str:
+    return """
+<script>
+const unitSearch = document.getElementById('unit-search');
+const splitOnly = document.getElementById('split-only');
+const unitCards = Array.from(document.querySelectorAll('.execution-unit'));
+const visibleCount = document.getElementById('visible-unit-count');
+function filterUnits() {
+  const query = unitSearch.value.trim().toLowerCase();
+  let visible = 0;
+  unitCards.forEach((card) => {
+    const matchesText = !query || card.dataset.search.includes(query);
+    const matchesSplit = !splitOnly.checked || card.dataset.split === 'true';
+    card.hidden = !(matchesText && matchesSplit);
+    if (!card.hidden) visible += 1;
+  });
+  visibleCount.textContent = `${visible.toLocaleString()} of ${unitCards.length.toLocaleString()} units`;
+}
+unitSearch.addEventListener('input', filterUnits);
+splitOnly.addEventListener('change', filterUnits);
+filterUnits();
+</script>
+"""
+
+
+def _render_execution_proposal_html(
+    execution_units: Sequence[dict[str, Any]],
+    category_execution: Sequence[dict[str, Any]],
+    planned_total: int,
+) -> str:
+    def render_metrics(items: Sequence[tuple[str, str]]) -> str:
+        return (
+            '<div class="execution-metrics">'
+            + "".join(
+                '<div class="execution-metric">'
+                f'<span class="metric-label">{html.escape(label)}</span>'
+                f'<span class="metric-value">{html.escape(value)}</span></div>'
+                for label, value in items
+            )
+            + "</div>"
+        )
+
+    ordered_units = sorted(
+        execution_units,
+        key=lambda row: (int(row["estimated_peak_local_bytes"]), row["unit_id"]),
+        reverse=True,
+    )
+    peak_values = [int(row["estimated_peak_local_bytes"]) for row in ordered_units]
+    units_by_worker: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in ordered_units:
+        units_by_worker[str(row["worker_instance_type"])].append(row)
+    split_categories = sorted(
+        (row for row in category_execution if int(row["execution_unit_count"]) > 1),
+        key=lambda row: (
+            int(row["execution_unit_count"]),
+            int(row["largest_estimated_peak_local_bytes"]),
+            row["leaf_id"],
+        ),
+        reverse=True,
+    )
+    planned_output_shards = sum(int(row["planned_output_shard_count"]) for row in ordered_units)
+    average_output_shard_bytes = [
+        int(row["average_output_shard_bytes"])
+        for row in ordered_units
+        for _ in range(int(row["planned_output_shard_count"]))
+    ]
+    metrics = render_metrics(
+        [
+            ("Execution units", f"{len(ordered_units):,}"),
+            ("Output shards", f"{planned_output_shards:,}"),
+            ("Output files", f"{planned_output_shards * 2:,}"),
+            ("Worker types", f"{len(units_by_worker):,}"),
+            ("Planned output", f"{_human_token_count(planned_total)} tokens"),
+        ]
+    )
+    storage_metrics = render_metrics(
+        [
+            ("Median unit", _human_byte_count(_percentile(peak_values, 0.5))),
+            ("P95 unit", _human_byte_count(_percentile(peak_values, 0.95))),
+            ("Largest unit", _human_byte_count(max(peak_values, default=0))),
+            (
+                "Median output shard",
+                _human_byte_count(_percentile(average_output_shard_bytes, 0.5)),
+            ),
+            (
+                "P95 output shard",
+                _human_byte_count(_percentile(average_output_shard_bytes, 0.95)),
+            ),
+        ]
+    )
+    worker_grid_html = (
+        '<table class="worker-grid"><thead><tr><th>Instance</th><th>Units</th>'
+        "<th>vCPUs</th><th>Local NVMe</th><th>Layout</th>"
+        "<th>Largest working set</th><th>Largest work estimate</th></tr></thead><tbody>"
+        + "".join(
+            "<tr>"
+            f"<td>{html.escape(instance_type)}</td>"
+            f"<td>{len(rows):,}</td>"
+            f"<td>{int(rows[0]['worker_vcpus']):,}</td>"
+            f"<td>{_human_byte_count(int(rows[0]['worker_local_nvme_bytes']))}</td>"
+            f"<td>{html.escape(str(rows[0]['worker_storage_layout']))}</td>"
+            f"<td>{_human_byte_count(max(int(row['estimated_peak_local_bytes']) for row in rows))}</td>"
+            f"<td>{_human_token_count(max(int(row['estimated_work_uint32_values']) for row in rows))}</td>"
+            "</tr>"
+            for instance_type, rows in sorted(
+                units_by_worker.items(), key=lambda item: int(item[1][0]["worker_vcpus"])
+            )
+        )
+        + "</tbody></table>"
+    )
+    split_html = (
+        '<div class="split-list">'
+        + "".join(
+            '<article class="split-category">'
+            f'<span class="split-name">{html.escape(str(row["mix_name"]))} / '
+            f"{html.escape(str(row['category_name']))}</span>"
+            f'<span class="split-value"><span>Units</span>{int(row["execution_unit_count"]):,}</span>'
+            f'<span class="split-value"><span>Output shards</span>'
+            f'{int(row["planned_output_shard_count"]):,}</span>'
+            f'<span class="split-value"><span>Output</span>'
+            f"{_human_token_count(int(row['planned_uint32_values']))} tokens</span>"
+            f'<span class="split-value"><span>Largest unit</span>'
+            f"{_human_byte_count(int(row['largest_estimated_peak_local_bytes']))}</span>"
+            "</article>"
+            for row in split_categories
+        )
+        + "</div>"
+        if split_categories
+        else '<p class="storage-note">No category requires more than one execution unit.</p>'
+    )
+    unit_cards: list[str] = []
+    for row in ordered_units:
+        input_bytes = int(row["input_npy_bytes"]) + int(row["input_metadata_bytes"])
+        output_bytes = int(row["output_npy_bytes"]) + int(row["estimated_output_metadata_bytes"])
+        selection_bytes = int(row["estimated_selection_index_bytes"])
+        peak_bytes = int(row["estimated_peak_local_bytes"])
+        worker_disk_bytes = int(row["worker_local_nvme_bytes"])
+        utilization = peak_bytes / worker_disk_bytes
+        source_width = 100 * input_bytes / worker_disk_bytes
+        output_width = 100 * output_bytes / worker_disk_bytes
+        selection_width = 100 * selection_bytes / worker_disk_bytes
+        category_label = f"{row['mix_name']} / {row['category_name']}"
+        searchable = html.escape(
+            f"{row['unit_id']} {row['mix_name']} {row['category_name']} " f"{row['worker_instance_type']}".lower(),
+            quote=True,
+        )
+        split = int(row["unit_count_for_category"]) > 1
+        unit_cards.append(
+            f'<details class="execution-unit" data-search="{searchable}" '
+            f'data-split="{str(split).lower()}"><summary>'
+            f'<span class="unit-title">{html.escape(category_label)}'
+            f'<span class="unit-position">Unit {int(row["unit_index"]):,} of '
+            f"{int(row['unit_count_for_category']):,}</span></span>"
+            f'<span class="unit-disk"><strong>{html.escape(str(row["worker_instance_type"]))}</strong>'
+            f"<span>{_human_byte_count(peak_bytes)} working set · "
+            f"{utilization:.1%} of local NVMe</span></span></summary>"
+            '<div class="unit-body"><div class="unit-metrics">'
+            '<div class="unit-metric"><span>Output tokens</span>'
+            f"<strong>{_human_token_count(int(row['planned_uint32_values']))}</strong></div>"
+            '<div class="unit-metric"><span>Estimated processing work</span>'
+            f"<strong>{_human_token_count(int(row['estimated_work_uint32_values']))}</strong></div>"
+            '<div class="unit-metric"><span>Worker</span>'
+            f"<strong>{int(row['worker_vcpus']):,} vCPUs · "
+            f"{_human_byte_count(worker_disk_bytes)}</strong></div>"
+            '<div class="unit-metric"><span>Reshard concurrency</span>'
+            f"<strong>{int(row['max_workers']):,} workers</strong></div>"
+            '<div class="unit-metric"><span>Output layout</span>'
+            f"<strong>{int(row['planned_output_shard_count']):,} shards · "
+            f"{_human_byte_count(int(row['average_output_shard_bytes']))} average</strong></div>"
+            '<div class="unit-metric"><span>Source shard downloads</span>'
+            f"<strong>{int(row['unique_object_count']):,}</strong></div>"
+            '<div class="unit-metric"><span>Source shards using document selection</span>'
+            f"<strong>{int(row['partial_object_count']):,}</strong></div>"
+            "</div>"
+            '<div class="disk-breakdown" aria-label="Estimated local disk composition">'
+            f'<span class="disk-segment disk-source" style="width:{source_width:.8f}%"></span>'
+            f'<span class="disk-segment disk-output" style="width:{output_width:.8f}%"></span>'
+            f'<span class="disk-segment disk-selection" style="width:{selection_width:.8f}%"></span>'
+            "</div>"
+            '<div class="disk-legend">'
+            f'<span class="legend-source">Source download: {_human_byte_count(input_bytes)}</span>'
+            f'<span class="legend-output">Materialized output: {_human_byte_count(output_bytes)}</span>'
+            f'<span class="legend-selection">Selection indexes: {_human_byte_count(selection_bytes)}</span>'
+            "</div>"
+            '<dl class="unit-paths"><div><dt>Unit ID</dt>'
+            f"<dd><code>{html.escape(str(row['unit_id']))}</code></dd></div>"
+            "<div><dt>Destination</dt>"
+            f"<dd><code>{html.escape(str(row['destination_prefix']))}</code></dd></div></dl>"
+            '<nav class="artifact-links" aria-label="Execution-unit artifacts">'
+            f'<a href="{_proposal_artifact_href(row["config_path"])}">Config</a>'
+            f'<a href="{_proposal_artifact_href(row["manifest_path"])}">Manifest</a>'
+            f'<a href="{_proposal_artifact_href(row["launcher_path"])}">Launcher</a>'
+            "</nav></div></details>"
+        )
+    return (
+        '<!doctype html><html><head><meta charset="utf-8">'
+        '<meta name="viewport" content="width=device-width,initial-scale=1">'
+        "<title>Dolma 3.5 materialization execution proposal</title>"
+        + _execution_proposal_style()
+        + "</head><body><h1>Dolma 3.5 Materialization Execution Proposal</h1>"
+        '<p class="lede">Worker partitioning, local-disk requirements, and runnable artifacts for the '
+        "materialization phase.</p>"
+        + metrics
+        + "<h2>Planned worker fleet</h2>"
+        + storage_metrics
+        + worker_grid_html
+        + "<h2>Categories requiring multiple execution units</h2>"
+        + split_html
+        + '<div class="unit-heading"><h2>Execution units</h2>'
+        f'<span class="visible-count" id="visible-unit-count">{len(ordered_units):,} of '
+        f"{len(ordered_units):,} units</span></div>"
+        '<div class="unit-controls"><input id="unit-search" type="search" '
+        'placeholder="Filter by source, category, or unit ID" aria-label="Filter execution units">'
+        '<label><input id="split-only" type="checkbox"> Only categories with multiple units</label></div>'
+        f'<div class="unit-list">{"".join(unit_cards)}</div>' + _execution_proposal_script() + "</body></html>\n"
+    )
+
+
+def _report_document_with_base(document: str, relative_base: str) -> str:
+    marker = "<head>"
+    if marker not in document:
+        raise PreparationError("Generated report is missing its HTML head")
+    return document.replace(
+        marker, f'<head><base href="{html.escape(relative_base, quote=True)}">', 1
+    )
+
+
+def _combine_plan_reports(build: Path) -> None:
+    """Write 01-plan/report.html with the inventory and execution reports embedded.
+
+    Each stage report is inlined in a template element, then the stage copies
+    are deleted.
+    """
+
+    plan_root = build / "01-plan"
+    resolution_report_path = plan_root / "resolution/report.html"
+    inventory_report_path = plan_root / "inventory/report.html"
+    execution_report_path = plan_root / "execution/report.html"
+    for report_path in (inventory_report_path, execution_report_path):
+        if report_path.is_symlink() or not report_path.is_file():
+            raise PreparationError(
+                f"Cannot compose missing or unsafe report: {report_path}"
+            )
+
+    inventory_document = _report_document_with_base(
+        inventory_report_path.read_text(encoding="utf-8"),
+        "inventory/",
+    )
+    execution_document = _report_document_with_base(
+        execution_report_path.read_text(encoding="utf-8"),
+        "execution/",
+    )
+    shell_palette = report_root_style(
+        (
+            ("shell", "shell"),
+            ("tab", "shell-tab"),
+            ("selected", "accent"),
+            ("muted", "muted"),
+            ("line", "line"),
+        )
+    )
+    combined = f"""<!doctype html>
+<html>
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width,initial-scale=1">
+<title>Dolma 3.5 Resharding Plan</title>
+<style>
+{shell_palette}
+*{{box-sizing:border-box}}html,body{{height:100%;margin:0}}body{{overflow:hidden;background:var(--shell);color:CanvasText;font:14px/1.4 system-ui,sans-serif}}
+.report-shell{{display:grid;height:100%;grid-template-rows:auto minmax(0,1fr)}}
+.report-tabs{{display:flex;gap:8px;padding:10px 18px;border-bottom:1px solid var(--line);background:var(--shell)}}
+.report-tab{{padding:9px 14px;border:0;border-radius:7px;background:transparent;color:var(--muted);font:inherit;font-weight:650;cursor:pointer}}
+.report-tab:hover{{background:var(--tab);color:CanvasText}}.report-tab[aria-selected="true"]{{background:var(--tab);color:var(--selected)}}
+.report-panel{{min-height:0}}.report-panel[hidden]{{display:none}}.report-frame{{display:block;width:100%;height:100%;border:0;background:Canvas}}
+@media(max-width:620px){{.report-tabs{{padding:8px}}.report-tab{{flex:1;padding:9px 8px}}}}
+</style>
+</head>
+<body>
+<template id="source-report-document">{inventory_document}</template>
+<template id="execution-report-document">{execution_document}</template>
+<main class="report-shell">
+  <nav class="report-tabs" role="tablist" aria-label="Plan report views">
+    <button class="report-tab" id="source-report-tab" type="button" role="tab" aria-controls="source-report-panel" aria-selected="true" data-report="source">Source inventory &amp; sampling</button>
+    <button class="report-tab" id="execution-report-tab" type="button" role="tab" aria-controls="execution-report-panel" aria-selected="false" data-report="execution">Materialization execution</button>
+  </nav>
+  <section class="report-panel" id="source-report-panel" role="tabpanel" aria-labelledby="source-report-tab">
+    <iframe class="report-frame" id="source-report-frame" title="Source inventory and sampling report"></iframe>
+  </section>
+  <section class="report-panel" id="execution-report-panel" role="tabpanel" aria-labelledby="execution-report-tab" hidden>
+    <iframe class="report-frame" id="execution-report-frame" title="Materialization execution report"></iframe>
+  </section>
+</main>
+<script>
+const reportNames = ['source', 'execution'];
+reportNames.forEach((name) => {{
+  const template = document.getElementById(`${{name}}-report-document`);
+  document.getElementById(`${{name}}-report-frame`).srcdoc = template.innerHTML;
+}});
+const reportTabs = Array.from(document.querySelectorAll('.report-tab'));
+function selectReport(name, updateHash = true) {{
+  reportTabs.forEach((tab) => {{
+    const selected = tab.dataset.report === name;
+    tab.setAttribute('aria-selected', String(selected));
+    tab.tabIndex = selected ? 0 : -1;
+    document.getElementById(`${{tab.dataset.report}}-report-panel`).hidden = !selected;
+  }});
+  if (updateHash) history.replaceState(null, '', `#${{name}}`);
+}}
+reportTabs.forEach((tab, index) => {{
+  tab.addEventListener('click', () => selectReport(tab.dataset.report));
+  tab.addEventListener('keydown', (event) => {{
+    if (!['ArrowLeft', 'ArrowRight'].includes(event.key)) return;
+    event.preventDefault();
+    const offset = event.key === 'ArrowRight' ? 1 : -1;
+    const next = reportTabs[(index + offset + reportTabs.length) % reportTabs.length];
+    selectReport(next.dataset.report);
+    next.focus();
+  }});
+}});
+selectReport(location.hash === '#execution' ? 'execution' : 'source', false);
+</script>
+</body>
+</html>
+"""
+    _write_text(plan_root / "report.html", combined)
+    for stage_report_path in (
+        resolution_report_path,
+        inventory_report_path,
+        execution_report_path,
+    ):
+        if stage_report_path.exists():
+            if stage_report_path.is_symlink() or not stage_report_path.is_file():
+                raise PreparationError(
+                    f"Refusing to remove an unsafe stage report: {stage_report_path}"
+                )
+            stage_report_path.unlink()
+
+
+def _percent_of(part: int, whole: int) -> float:
+    """Return part as a percentage of whole, treating an empty whole as zero."""
+
+    return 100 * part / whole if whole else 0.0
+
+
+def _token_share_metric(
+    label: str, value: int, percent: float, suffix: str = ""
+) -> dict[str, str]:
+    """Return one metric column: a token count and its share of a parent total."""
+
+    share = f"{percent:.2f}%" + (f" {suffix}" if suffix else "")
+    return {"label": label, "value": f"{_human_token_count(value)} tokens · {share}"}
+
+
+def _group_subcategories_by_family(
+    rows: Sequence[dict[str, Any]],
+) -> dict[str, list[dict[str, Any]]]:
+    """Group mix rows by source family.
+
+    Each row is annotated in place with source_family, subcategory_name, and
+    display_name; the chart and detail renderers read the same dictionaries.
+    """
+
+    grouped: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in rows:
+        source_family, subcategory = _split_mix_name(str(row["mix_name"]))
+        row["source_family"] = source_family
+        row["subcategory_name"] = subcategory
+        row["display_name"] = subcategory
+        grouped[source_family].append(row)
+    return grouped
+
+
+def _family_chart_rows(
+    family_rows: Sequence[dict[str, Any]],
+    *,
+    value_field: str,
+    percent_field: str,
+    family_percent_field: str,
+    detail_prefix: str,
+    subcategory_prefix: str,
+    total: int,
+) -> str:
+    """Render the source-family chart and each family's subcategory chart.
+
+    Returns the family chart rows and stores each family's subcategory chart
+    under its subcategory_chart_rows key. Both levels gain accordion IDs here.
+    """
+
+    chart_rows = _interactive_chart_rows(
+        family_rows, value_field, percent_field, detail_prefix, total
+    )
+    for family_index, family_row in enumerate(family_rows):
+        family_row["subcategory_chart_rows"] = _interactive_chart_rows(
+            family_row["subcategories"],
+            value_field,
+            family_percent_field,
+            f"{subcategory_prefix}-{family_index}",
+            int(family_row[value_field]),
+            row_class="subcategory-row",
+        )
+    return chart_rows
+
+
+def _subcategory_detail_section(
+    row: dict[str, Any], total_html: str, category_sections: Sequence[str]
+) -> str:
+    """Wrap one subcategory's category grid in its collapsed accordion panel."""
+
+    return (
+        f'<section class="subcategory-detail" id="{row["detail_id"]}" hidden>'
+        '<div class="detail-head">'
+        f"<h2>{html.escape(str(row['subcategory_name']))}</h2>"
+        f'<div class="detail-total">{total_html}</div></div>'
+        '<div class="category-grid">' + "".join(category_sections) + "</div></section>"
+    )
+
+
+def _source_to_target_total_html(row: dict[str, Any], source: int, target: int) -> str:
+    """Render one inventory accordion's source-to-target totals and sampling rate."""
+
+    return (
+        f"source {_human_token_count(source)} → "
+        f"target {_human_token_count(target)}<br>"
+        f'<span class="sampling {row["sampling_class"]}">'
+        f"{html.escape(str(row['sampling_rate']))}</span>"
+    )
+
+
+def _family_detail_sections(
+    family_rows: Sequence[dict[str, Any]],
+    subcategory_detail_by_mix: dict[str, str],
+    render_total: Callable[[dict[str, Any]], str],
+) -> str:
+    """Assemble one collapsed accordion panel per source family."""
+
+    return "".join(
+        f'<section class="mix-detail" id="{family_row["detail_id"]}" hidden>'
+        '<div class="detail-head">'
+        f"<h2>{html.escape(str(family_row['mix_name']))}</h2>"
+        f'<div class="detail-total">{render_total(family_row)}</div></div>'
+        '<div class="subcategory-list">'
+        + str(family_row["subcategory_chart_rows"])
+        + "</div>"
+        + "".join(
+            subcategory_detail_by_mix[str(subcategory["mix_name"])]
+            for subcategory in family_row["subcategories"]
+        )
+        + "</section>"
+        for family_row in family_rows
+    )
+
+
+def _render_plan_report(
+    phase: Path,
+    normalized_mix: Sequence[dict[str, Any]],
+    normalized_paths: Sequence[dict[str, Any]],
+    catalog_matches: Sequence[dict[str, Any]],
+    direct_patterns: Sequence[dict[str, Any]],
+) -> None:
+    plots = phase / "plots"
+    plot_data = phase / "plot-data"
+    plots.mkdir(exist_ok=False)
+    plot_data.mkdir(exist_ok=False)
+    target_by_mix: dict[str, int] = defaultdict(int)
+    for row in normalized_mix:
+        target_by_mix[row["mix_name"]] += int(row["target_uint32_values"])
+    target_total = sum(target_by_mix.values())
+    target_rows = [
+        {
+            "mix_name": name,
+            "target_uint32_values": value,
+            "target_percent": f"{100 * value / target_total:.8f}"
+            if target_total
+            else "0",
+        }
+        for name, value in sorted(
+            target_by_mix.items(), key=lambda item: item[1], reverse=True
+        )
+    ]
+    _write_csv(
+        plot_data / "target-mix.csv",
+        target_rows,
+        ["mix_name", "target_uint32_values", "target_percent"],
+    )
+    family_rows: list[dict[str, Any]] = []
+    for source_family, subcategories in _group_subcategories_by_family(
+        target_rows
+    ).items():
+        family_target = sum(int(row["target_uint32_values"]) for row in subcategories)
+        family_percent = _percent_of(family_target, target_total)
+        ordered_subcategories = sorted(
+            subcategories,
+            key=lambda row: (
+                int(row["target_uint32_values"]),
+                str(row["subcategory_name"]),
+            ),
+            reverse=True,
+        )
+        for row in ordered_subcategories:
+            sub_target = int(row["target_uint32_values"])
+            sub_percent = _percent_of(sub_target, family_target)
+            row["family_target_percent"] = f"{sub_percent:.8f}"
+            row["metric_columns"] = [
+                _token_share_metric("Target", sub_target, sub_percent, "of family")
+            ]
+        family_rows.append(
+            {
+                "mix_name": source_family,
+                "target_uint32_values": family_target,
+                "target_percent": f"{family_percent:.8f}",
+                "subcategories": ordered_subcategories,
+                "metric_columns": [
+                    _token_share_metric("Target", family_target, family_percent)
+                ],
+            }
+        )
+    family_rows.sort(
+        key=lambda row: (
+            int(row["target_uint32_values"]),
+            str(row["mix_name"]),
+        ),
+        reverse=True,
+    )
+    _write_text(
+        plots / "target-mix.svg",
+        _svg_bar_chart(
+            "Target allocation by source family",
+            [row["mix_name"] for row in family_rows],
+            [float(row["target_percent"]) for row in family_rows],
+            "% of target",
+            value_labels=[
+                _mix_plot_value_label(row["target_uint32_values"], target_total)
+                for row in family_rows
+            ],
+            summary=(
+                f"Total target: {_human_token_count(target_total)} tokens "
+                f"({target_total:,})"
+            ),
+        ),
+    )
+    chart_rows = _family_chart_rows(
+        family_rows,
+        value_field="target_uint32_values",
+        percent_field="target_percent",
+        family_percent_field="family_target_percent",
+        detail_prefix="plan-family-detail",
+        subcategory_prefix="plan-subcategory",
+        total=target_total,
+    )
+    paths_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in normalized_paths:
+        paths_by_leaf[row["leaf_id"]].append(row)
+    catalog_counts = Counter(row["path_id"] for row in catalog_matches)
+    catalog_sources_by_path: dict[str, list[str]] = defaultdict(list)
+    for row in catalog_matches:
+        catalog_sources_by_path[row["path_id"]].append(
+            f"s3://{row['bucket']}/{row['key']}"
+        )
+    direct_sources_by_path = {
+        row["path_id"]: f"s3://{row['bucket']}/{row['key_pattern']}"
+        for row in direct_patterns
+    }
+    categories_by_mix: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in normalized_mix:
+        categories_by_mix[row["mix_name"]].append(row)
+
+    subcategory_detail_by_mix: dict[str, str] = {}
+    for target_row in target_rows:
+        mix_name = str(target_row["mix_name"])
+        mix_target = int(target_row["target_uint32_values"])
+        category_sections: list[str] = []
+        for category in sorted(
+            categories_by_mix[mix_name], key=lambda row: int(row["category_index"])
+        ):
+            category_target = int(category["target_uint32_values"])
+            category_percent = 100 * category_target / mix_target if mix_target else 0
+            path_rows: list[str] = []
+            for path in sorted(
+                paths_by_leaf[category["leaf_id"]], key=lambda row: row["path_id"]
+            ):
+                active = path["active"] == "true"
+                source_uris = sorted(set(catalog_sources_by_path[path["path_id"]]))
+                direct_source = direct_sources_by_path.get(path["path_id"])
+                if direct_source is not None:
+                    source_uris.append(direct_source)
+                if not active:
+                    matched = "inactive"
+                elif direct_source is not None:
+                    matched = "pending inventory"
+                else:
+                    matched = _count_label(
+                        catalog_counts[path["path_id"]], "matched NPY"
+                    )
+                source_details = _source_uri_details(
+                    source_uris, "No resolved S3 source"
+                )
+                path_rows.append(
+                    '<details class="path-detail"><summary>'
+                    f'<span class="path-name">{html.escape(_path_subgroup(path["yaml_path"]))}</span>'
+                    f'<span class="path-stat">{html.escape(matched)}'
+                    '<span class="path-chevron" aria-hidden="true">›</span></span>'
+                    "</summary>"
+                    f"<code>{source_details}</code>"
+                    "</details>"
+                )
+            category_sections.append(
+                '<section class="category">'
+                '<div class="category-head">'
+                f'<span class="category-name">{html.escape(category["category_name"])}</span>'
+                f'<span class="category-value">{html.escape(_human_token_count(category_target))} tokens · {category_percent:.2f}%</span>'
+                "</div>"
+                '<div class="category-bar" aria-hidden="true">'
+                f'<span class="category-bar-fill" style="width:{category_percent:.8f}%"></span></div>'
+                '<div class="path-list">' + "".join(path_rows) + "</div>" + "</section>"
+            )
+        subcategory_detail_by_mix[mix_name] = _subcategory_detail_section(
+            target_row,
+            f"{_human_token_count(mix_target)} tokens · "
+            f"{float(target_row['target_percent']):.2f}% of target",
+            category_sections,
+        )
+    detail_sections = _family_detail_sections(
+        family_rows,
+        subcategory_detail_by_mix,
+        lambda row: (
+            f"{_human_token_count(int(row['target_uint32_values']))} tokens · "
+            f"{float(row['target_percent']):.2f}% of target"
+        ),
+    )
+    report_html = (
+        '<!doctype html><html><head><meta charset="utf-8"><title>Dolma 3.5 target allocation and source path plan</title>'
+        + _interactive_report_style()
+        + "</head><body>"
+        + "<h1>Dolma 3.5 Target Allocation and Source Path Plan</h1>"
+        f'<div class="chart-total">Materialized output target: '
+        f"{_human_token_count(target_total)} tokens ({target_total:,})</div>"
+        f'<div class="mix-chart">{chart_rows}</div>'
+        + detail_sections
+        + _interactive_report_script()
+        + "</body></html>\n"
+    )
+    _write_text(phase / "report.html", report_html)
+
+
+def _build_inventory_details(
+    normalized_mix: Sequence[dict[str, Any]],
+    normalized_paths: Sequence[dict[str, Any]],
+    required_rows: Sequence[dict[str, Any]],
+) -> dict[str, Any]:
+    rows_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    rows_by_path: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    paths_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    categories_by_mix: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in required_rows:
+        rows_by_leaf[row["leaf_id"]].append(row)
+        rows_by_path[row["path_id"]].append(row)
+    for row in normalized_paths:
+        paths_by_leaf[row["leaf_id"]].append(row)
+    for row in normalized_mix:
+        categories_by_mix[row["mix_name"]].append(row)
+
+    source_rows: list[dict[str, Any]] = []
+    for mix_name, categories in categories_by_mix.items():
+        source_family, subcategory_name = _split_mix_name(str(mix_name))
+        category_rows: list[dict[str, Any]] = []
+        source_object_uris: set[str] = set()
+        for category in sorted(categories, key=lambda row: int(row["category_index"])):
+            leaf_id = category["leaf_id"]
+            category_objects = {
+                row["npy_uri"]: int(row["estimated_uint32_values"])
+                for row in rows_by_leaf[leaf_id]
+            }
+            source_object_uris.update(category_objects)
+            source_tokens = sum(category_objects.values())
+            target_tokens = int(category["target_uint32_values"])
+            path_definitions = sorted(
+                paths_by_leaf[leaf_id], key=lambda row: row["path_id"]
+            )
+            path_objects = [
+                {
+                    row["npy_uri"]: int(row["estimated_uint32_values"])
+                    for row in rows_by_path[path["path_id"]]
+                }
+                for path in path_definitions
+            ]
+            path_source_tokens = [sum(objects.values()) for objects in path_objects]
+            path_targets = _apportion_by_size(target_tokens, path_source_tokens)
+            lower_groups: list[dict[str, Any]] = []
+            for path, objects, path_source, path_target in zip(
+                path_definitions, path_objects, path_source_tokens, path_targets
+            ):
+                _, _, ratio = _sampling_change(path_source, path_target)
+                sampling_rate, _ = _sampling_rate_label(path_source, path_target)
+                lower_groups.append(
+                    {
+                        "path_id": path["path_id"],
+                        "lower_group": _path_subgroup(path["yaml_path"]),
+                        "active": path["active"] == "true",
+                        "yaml_path": path["yaml_path"],
+                        "source_uint32_values": path_source,
+                        "implied_target_uint32_values": path_target,
+                        "token_delta": path_target - path_source,
+                        "sampling_ratio": ratio,
+                        "sampling_rate": sampling_rate,
+                        "unique_npy_count": len(objects),
+                    }
+                )
+            _, _, ratio = _sampling_change(source_tokens, target_tokens)
+            sampling_rate, _ = _sampling_rate_label(source_tokens, target_tokens)
+            category_rows.append(
+                {
+                    "leaf_id": leaf_id,
+                    "category_name": category["category_name"],
+                    "active": category["active"] == "true",
+                    "source_uint32_values": source_tokens,
+                    "target_uint32_values": target_tokens,
+                    "token_delta": target_tokens - source_tokens,
+                    "sampling_ratio": ratio,
+                    "sampling_rate": sampling_rate,
+                    "unique_npy_count": len(category_objects),
+                    "lower_groups": lower_groups,
+                }
+            )
+        source_tokens = sum(row["source_uint32_values"] for row in category_rows)
+        target_tokens = sum(row["target_uint32_values"] for row in category_rows)
+        for category in category_rows:
+            category["source_percent_of_parent"] = (
+                100 * category["source_uint32_values"] / source_tokens
+                if source_tokens
+                else 0.0
+            )
+            category["target_percent_of_parent"] = (
+                100 * category["target_uint32_values"] / target_tokens
+                if target_tokens
+                else 0.0
+            )
+            for lower_group in category["lower_groups"]:
+                lower_group["source_percent_of_parent"] = (
+                    100
+                    * lower_group["source_uint32_values"]
+                    / category["source_uint32_values"]
+                    if category["source_uint32_values"]
+                    else 0.0
+                )
+                lower_group["implied_target_percent_of_parent"] = (
+                    100
+                    * lower_group["implied_target_uint32_values"]
+                    / category["target_uint32_values"]
+                    if category["target_uint32_values"]
+                    else 0.0
+                )
+        _, _, ratio = _sampling_change(source_tokens, target_tokens)
+        sampling_rate, _ = _sampling_rate_label(source_tokens, target_tokens)
+        source_rows.append(
+            {
+                "mix_name": mix_name,
+                "source_family": source_family,
+                "subcategory_name": subcategory_name,
+                "source_uint32_values": source_tokens,
+                "target_uint32_values": target_tokens,
+                "token_delta": target_tokens - source_tokens,
+                "sampling_ratio": ratio,
+                "sampling_rate": sampling_rate,
+                "unique_npy_count": len(source_object_uris),
+                "categories": category_rows,
+            }
+        )
+
+    source_total = sum(row["source_uint32_values"] for row in source_rows)
+    target_total = sum(row["target_uint32_values"] for row in source_rows)
+    for source in source_rows:
+        source["source_percent_of_total"] = (
+            100 * source["source_uint32_values"] / source_total if source_total else 0.0
+        )
+        source["target_percent_of_total"] = (
+            100 * source["target_uint32_values"] / target_total if target_total else 0.0
+        )
+        for category in source["categories"]:
+            category["source_percent_of_total"] = (
+                100 * category["source_uint32_values"] / source_total
+                if source_total
+                else 0.0
+            )
+            category["target_percent_of_total"] = (
+                100 * category["target_uint32_values"] / target_total
+                if target_total
+                else 0.0
+            )
+            for lower_group in category["lower_groups"]:
+                lower_group["source_percent_of_total"] = (
+                    100 * lower_group["source_uint32_values"] / source_total
+                    if source_total
+                    else 0.0
+                )
+                lower_group["implied_target_percent_of_total"] = (
+                    100 * lower_group["implied_target_uint32_values"] / target_total
+                    if target_total
+                    else 0.0
+                )
+
+    source_rows.sort(
+        key=lambda row: (
+            row["target_uint32_values"],
+            row["source_uint32_values"],
+            row["mix_name"],
+        ),
+        reverse=True,
+    )
+    _, _, aggregate_ratio = _sampling_change(source_total, target_total)
+    aggregate_rate, _ = _sampling_rate_label(source_total, target_total)
+    return {
+        "schema_version": 1,
+        "source_uint32_values": source_total,
+        "target_uint32_values": target_total,
+        "token_delta": target_total - source_total,
+        "sampling_ratio": aggregate_ratio,
+        "sampling_rate": aggregate_rate,
+        "source_family_count": len({str(row["source_family"]) for row in source_rows}),
+        "subcategory_count": len(source_rows),
+        "source_count": len(source_rows),
+        "category_count": sum(len(row["categories"]) for row in source_rows),
+        "lower_group_count": sum(
+            len(category["lower_groups"])
+            for source in source_rows
+            for category in source["categories"]
+        ),
+        "sources": source_rows,
+    }
+
+
+def _replace_inventory_json(build: Path, path: Path, value: Any) -> None:
+    _validate_preparation_build(build)
+    phase = build / "01-plan/inventory"
+    if path.parent != phase or path.name not in {
+        "inventory-summary.json",
+        "inventory-details.json",
+    }:
+        raise PreparationError(f"Refusing to replace non-inventory artifact: {path}")
+    if path.exists() and (path.is_symlink() or not path.is_file()):
+        raise PreparationError(f"Unsafe generated inventory artifact: {path}")
+    temporary = path.with_name(f".{path.name}.{os.getpid()}.tmp")
+    _write_json(temporary, value)
+    try:
+        os.replace(temporary, path)
+    except Exception:
+        if temporary.exists():
+            temporary.unlink()
+        raise
+
+
+# Not called by any entry point; exercised only by tests/python/test_dolma3p5_resharding.py.
+def refresh_inventory_details(build: Path) -> dict[str, Any]:
+    build = build.resolve()
+    _validate_preparation_build(build)
+    phase = build / "01-plan/inventory"
+    summary_path = phase / "inventory-summary.json"
+    if not summary_path.is_file() or summary_path.is_symlink():
+        raise PreparationError(f"Inventory summary is missing: {summary_path}")
+    with summary_path.open(encoding="utf-8") as f:
+        summary = json.load(f)
+    details = _build_inventory_details(
+        normalized_mix=_read_csv(build / "01-plan/resolution/normalized-mix.csv"),
+        normalized_paths=_read_csv(build / "01-plan/resolution/normalized-paths.csv"),
+        required_rows=_read_csv(phase / "required-objects.csv"),
+    )
+    summary_source = int(
+        summary["source_uint32_values"]
+        if "source_uint32_values" in summary
+        else summary["original_uint32_values"]
+    )
+    if summary_source != details["source_uint32_values"]:
+        raise PreparationError("Inventory detail source total does not match summary")
+    if int(summary["target_uint32_values"]) != details["target_uint32_values"]:
+        raise PreparationError("Inventory detail target total does not match summary")
+    metadata = {
+        "source_uint32_values": details["source_uint32_values"],
+        "token_delta": details["token_delta"],
+        "sampling_ratio": details["sampling_ratio"],
+        "sampling_rate": details["sampling_rate"],
+        "source_family_count": details["source_family_count"],
+        "subcategory_count": details["subcategory_count"],
+        "source_count": details["source_count"],
+        "category_count": details["category_count"],
+        "lower_group_count": details["lower_group_count"],
+        "details_artifact": "inventory-details.json",
+        "report_artifact": "../report.html",
+    }
+    summary.pop("sampling_change", None)
+    summary.update(metadata)
+    _replace_inventory_json(build, phase / "inventory-details.json", details)
+    _replace_inventory_json(build, summary_path, summary)
+    return summary
+
+
+def _inventory_path_details(
+    path_definitions: Sequence[dict[str, Any]],
+    path_objects: Sequence[dict[str, int]],
+    path_originals: Sequence[int],
+    path_targets: Sequence[int],
+    *,
+    leaf_id: str,
+    mix_name: str,
+    category_name: str,
+    category_original: int,
+    category_target: int,
+    path_comparisons: list[dict[str, Any]],
+) -> str:
+    """Render one category's YAML-path rows.
+
+    Appends one row per path to path_comparisons for
+    sampling-by-lower-group.csv. Returns the collapsed path list's HTML.
+    """
+
+    path_rows: list[str] = []
+    for path, objects, path_original, path_target in zip(
+        path_definitions, path_objects, path_originals, path_targets
+    ):
+        _, path_class, path_ratio = _sampling_change(path_original, path_target)
+        path_sampling_rate, _ = _sampling_rate_label(path_original, path_target)
+        path_source_percent = _percent_of(path_original, category_original)
+        path_target_percent = _percent_of(path_target, category_target)
+        path_comparisons.append(
+            {
+                "path_id": path["path_id"],
+                "leaf_id": leaf_id,
+                "mix_name": mix_name,
+                "category_name": category_name,
+                "lower_group": _path_subgroup(path["yaml_path"]),
+                "yaml_path": path["yaml_path"],
+                "original_uint32_values": path_original,
+                "source_percent_of_parent": path_source_percent,
+                "implied_target_uint32_values": path_target,
+                "implied_target_percent_of_parent": path_target_percent,
+                "sampling_ratio": "" if path_ratio is None else f"{path_ratio:.12g}",
+                "sampling_rate": path_sampling_rate,
+                "unique_npy_count": len(objects),
+            }
+        )
+        source_details = _source_uri_details(objects, "No source NPYs resolved")
+        path_rows.append(
+            '<details class="path-detail"><summary>'
+            f'<span class="path-name">{html.escape(_path_subgroup(path["yaml_path"]))}</span>'
+            f'<span class="path-stat">{_count_label(len(objects), "file")}'
+            '<span class="path-chevron" aria-hidden="true">›</span></span>'
+            '<span class="path-sampling">'
+            '<span class="path-metric"><span class="mix-metric-label">Source</span>'
+            f'<span class="mix-metric-value">{_human_token_count(path_original)} tokens · '
+            f"{path_source_percent:.2f}% of category</span></span>"
+            '<span class="path-metric"><span class="mix-metric-label">Target</span>'
+            f'<span class="mix-metric-value">{_human_token_count(path_target)} tokens · '
+            f"{path_target_percent:.2f}% of category</span></span>"
+            '<span class="path-metric"><span class="mix-metric-label">Sampling</span>'
+            f'<span class="mix-metric-value sampling {path_class}">'
+            f"{html.escape(path_sampling_rate)}</span></span>"
+            + _comparison_bars(path_original, path_target)
+            + "</span></summary>"
+            '<code><span class="path-detail-metrics">'
+            f'<span class="path-detail-metric">Source: {path_original:,} tokens</span>'
+            f'<span class="path-detail-metric">Implied target: {path_target:,} tokens</span>'
+            f"</span>{source_details}</code>"
+            "</details>"
+        )
+    return "".join(path_rows)
+
+
+def _inventory_category_section(
+    category_name: str,
+    *,
+    source: int,
+    source_percent: float,
+    target: int,
+    target_percent: float,
+    sampling_rate: str,
+    sampling_class: str,
+    path_rows_html: str,
+) -> str:
+    """Render one category's source-versus-target block inside a subcategory panel."""
+
+    return (
+        '<section class="category">'
+        '<div class="category-head">'
+        f'<span class="category-name">{html.escape(category_name)}</span>'
+        "</div>"
+        '<div class="category-metrics">'
+        '<span class="category-metric"><span class="mix-metric-label">Source</span>'
+        f'<span class="mix-metric-value">{_human_token_count(source)} tokens · '
+        f"{source_percent:.2f}% of entry</span></span>"
+        '<span class="category-metric"><span class="mix-metric-label">Target</span>'
+        f'<span class="mix-metric-value">{_human_token_count(target)} tokens · '
+        f"{target_percent:.2f}% of entry</span></span>"
+        '<span class="category-metric"><span class="mix-metric-label">Sampling</span>'
+        f'<span class="mix-metric-value sampling {sampling_class}">'
+        f"{html.escape(sampling_rate)}</span></span></div>"
+        + _comparison_bars(source, target)
+        + '<div class="path-list">'
+        + path_rows_html
+        + "</div></section>"
+    )
+
+
+def _sampling_audit_html(sampling_rate_rows: Sequence[dict[str, Any]]) -> str:
+    """Render the banner for categories above maximum_expected_upsample_rate.
+
+    Returns an empty string when no category exceeds it.
+    """
+
+    failures = [
+        row for row in sampling_rate_rows if row["status"] == "above_expected_maximum"
+    ]
+    if not failures:
+        return ""
+    configured_maximum = failures[0]["maximum_expected_upsample_rate"]
+    return (
+        '<section class="audit-warning"><h2>Source-size consistency check failed</h2>'
+        f"<p>{len(failures):,} categories exceed the configured "
+        f"{float(configured_maximum):.2f}× expected maximum. Proposal generation is "
+        "blocked until the resolved source objects agree with the mix's source-size basis. "
+        "Exact values are in <code>sampling-rate-audit.csv</code>.</p><ul>"
+        + "".join(
+            "<li>"
+            f"{html.escape(str(row['mix_name']))} / "
+            f"{html.escape(str(row['category_name']))}: "
+            f"{float(row['sample_rate']):.2f}× "
+            f"({_human_token_count(int(row['original_uint32_values']))} original → "
+            f"{_human_token_count(int(row['target_uint32_values']))} target)"
+            "</li>"
+            for row in failures
+        )
+        + "</ul></section>"
+    )
+
+
+def _render_inventory_report(
+    phase: Path,
+    normalized_mix: Sequence[dict[str, Any]],
+    normalized_paths: Sequence[dict[str, Any]],
+    required_rows: Sequence[dict[str, Any]],
+    all_objects: Sequence[dict[str, Any]],
+    missing_rows: Sequence[dict[str, Any]],
+    resolution_failures: Sequence[dict[str, Any]],
+    invalid_sizes: Sequence[dict[str, Any]],
+    sampling_rate_rows: Sequence[dict[str, Any]],
+) -> dict[str, Any]:
+    plots = phase / "plots"
+    plot_data = phase / "plot-data"
+    plots.mkdir(exist_ok=False)
+    plot_data.mkdir(exist_ok=False)
+    found_npy = {(row["bucket"], row["key"]) for row in required_rows}
+    found_metadata = {
+        (row["bucket"], _pair_metadata_key(row["key"])) for row in required_rows
+    }
+    missing_npy = sum(row["object_type"] == "npy" for row in missing_rows)
+    missing_metadata = sum(row["object_type"] == "metadata" for row in missing_rows)
+    coverage_rows = [
+        {"state": "found NPYs", "count": len(found_npy)},
+        {"state": "missing NPYs", "count": missing_npy},
+        {"state": "found metadata", "count": len(found_metadata)},
+        {"state": "missing metadata", "count": missing_metadata},
+        {"state": "path resolution failures", "count": len(resolution_failures)},
+        {"state": "invalid NPY sizes", "count": len(invalid_sizes)},
+    ]
+    original_by_mix: dict[str, int] = defaultdict(int)
+    seen: set[tuple[str, str]] = set()
+    for row in required_rows:
+        identity = (row["leaf_id"], row["npy_uri"])
+        if identity in seen:
+            continue
+        seen.add(identity)
+        original_by_mix[row["mix_name"]] += int(row["estimated_uint32_values"])
+    target_by_mix: dict[str, int] = defaultdict(int)
+    categories_by_mix: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    target_by_leaf: dict[str, int] = {}
+    for row in normalized_mix:
+        target = int(row["target_uint32_values"])
+        target_by_mix[row["mix_name"]] += target
+        target_by_leaf[row["leaf_id"]] = target
+        categories_by_mix[row["mix_name"]].append(row)
+    original_total = sum(original_by_mix.values())
+    target_total = sum(target_by_mix.values())
+    mix_names = sorted(
+        set(original_by_mix) | set(target_by_mix),
+        key=lambda name: (target_by_mix[name], original_by_mix[name], name),
+        reverse=True,
+    )
+    comparison_rows: list[dict[str, Any]] = []
+    for name in mix_names:
+        original = original_by_mix[name]
+        target = target_by_mix[name]
+        source_percent = _percent_of(original, original_total)
+        target_percent = _percent_of(target, target_total)
+        _, change_class, ratio = _sampling_change(original, target)
+        sampling_rate, _ = _sampling_rate_label(original, target)
+        comparison_rows.append(
+            {
+                "mix_name": name,
+                "available_uint32_values": original,
+                "available_percent": f"{source_percent:.8f}",
+                "target_uint32_values": target,
+                "target_percent": f"{target_percent:.8f}",
+                "sampling_ratio": "" if ratio is None else f"{ratio:.12g}",
+                "sampling_rate": sampling_rate,
+                "sampling_class": change_class,
+                "metric_columns": [
+                    _token_share_metric("Source", original, source_percent),
+                    _token_share_metric("Target", target, target_percent),
+                    {"label": "Sampling", "value": sampling_rate},
+                ],
+            }
+        )
+
+    family_rows: list[dict[str, Any]] = []
+    for source_family, subcategories in _group_subcategories_by_family(
+        comparison_rows
+    ).items():
+        family_source = sum(
+            int(row["available_uint32_values"]) for row in subcategories
+        )
+        family_target = sum(int(row["target_uint32_values"]) for row in subcategories)
+        source_percent = _percent_of(family_source, original_total)
+        target_percent = _percent_of(family_target, target_total)
+        sampling_rate, sampling_class = _sampling_rate_label(
+            family_source, family_target
+        )
+        ordered_subcategories = sorted(
+            subcategories,
+            key=lambda row: (
+                int(row["target_uint32_values"]),
+                int(row["available_uint32_values"]),
+                str(row["subcategory_name"]),
+            ),
+            reverse=True,
+        )
+        for row in ordered_subcategories:
+            sub_source = int(row["available_uint32_values"])
+            sub_target = int(row["target_uint32_values"])
+            sub_source_percent = _percent_of(sub_source, family_source)
+            sub_target_percent = _percent_of(sub_target, family_target)
+            row["family_target_percent"] = f"{sub_target_percent:.8f}"
+            row["metric_columns"] = [
+                _token_share_metric(
+                    "Source", sub_source, sub_source_percent, "of source"
+                ),
+                _token_share_metric(
+                    "Target", sub_target, sub_target_percent, "of target"
+                ),
+                {"label": "Sampling", "value": row["sampling_rate"]},
+            ]
+        family_rows.append(
+            {
+                "mix_name": source_family,
+                "available_uint32_values": family_source,
+                "available_percent": f"{source_percent:.8f}",
+                "target_uint32_values": family_target,
+                "target_percent": f"{target_percent:.8f}",
+                "sampling_rate": sampling_rate,
+                "sampling_class": sampling_class,
+                "subcategories": ordered_subcategories,
+                "metric_columns": [
+                    _token_share_metric("Source", family_source, source_percent),
+                    _token_share_metric("Target", family_target, target_percent),
+                    {"label": "Sampling", "value": sampling_rate},
+                ],
+            }
+        )
+    family_rows.sort(
+        key=lambda row: (
+            int(row["target_uint32_values"]),
+            int(row["available_uint32_values"]),
+            str(row["mix_name"]),
+        ),
+        reverse=True,
+    )
+    available_rows = sorted(
+        family_rows,
+        key=lambda row: int(row["available_uint32_values"]),
+        reverse=True,
+    )[:40]
+    _write_csv(plot_data / "coverage.csv", coverage_rows, ["state", "count"])
+    _write_csv(
+        plot_data / "available-by-mix.csv",
+        comparison_rows,
+        [
+            "mix_name",
+            "available_uint32_values",
+            "available_percent",
+            "target_uint32_values",
+            "target_percent",
+            "sampling_ratio",
+            "sampling_rate",
+        ],
+    )
+    _write_text(
+        plots / "coverage.svg",
+        _svg_bar_chart(
+            "Source inventory coverage",
+            [row["state"] for row in coverage_rows],
+            [row["count"] for row in coverage_rows],
+            "items",
+        ),
+    )
+    _write_text(
+        plots / "available-by-mix.svg",
+        _svg_bar_chart(
+            "Largest available mix entries by share of available tokens",
+            [row["mix_name"] for row in available_rows],
+            [float(row["available_percent"]) for row in available_rows],
+            "% of available tokens",
+            value_labels=[
+                _mix_plot_value_label(row["available_uint32_values"], original_total)
+                for row in available_rows
+            ],
+            summary=(
+                f"Source aggregate: {_human_token_count(original_total)} tokens "
+                f"({original_total:,})"
+            ),
+        ),
+    )
+    chart_rows = _family_chart_rows(
+        family_rows,
+        value_field="target_uint32_values",
+        percent_field="target_percent",
+        family_percent_field="family_target_percent",
+        detail_prefix="inventory-family-detail",
+        subcategory_prefix="inventory-subcategory",
+        total=target_total,
+    )
+    rows_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    rows_by_path: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in required_rows:
+        rows_by_leaf[row["leaf_id"]].append(row)
+        rows_by_path[row["path_id"]].append(row)
+    paths_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in normalized_paths:
+        paths_by_leaf[row["leaf_id"]].append(row)
+
+    category_comparisons: list[dict[str, Any]] = []
+    path_comparisons: list[dict[str, Any]] = []
+    subcategory_detail_by_mix: dict[str, str] = {}
+    for mix_row in comparison_rows:
+        mix_name = str(mix_row["mix_name"])
+        mix_original = int(mix_row["available_uint32_values"])
+        mix_target = int(mix_row["target_uint32_values"])
+        category_sections: list[str] = []
+        for category in sorted(
+            categories_by_mix[mix_name], key=lambda row: int(row["category_index"])
+        ):
+            leaf_id = category["leaf_id"]
+            leaf_rows = rows_by_leaf[leaf_id]
+            category_objects = {
+                row["npy_uri"]: int(row["estimated_uint32_values"]) for row in leaf_rows
+            }
+            category_original = sum(category_objects.values())
+            category_target = target_by_leaf[leaf_id]
+            category_source_percent = _percent_of(category_original, mix_original)
+            category_target_percent = _percent_of(category_target, mix_target)
+            _, category_class, category_ratio = _sampling_change(
+                category_original, category_target
+            )
+            category_sampling_rate, _ = _sampling_rate_label(
+                category_original, category_target
+            )
+            category_comparisons.append(
+                {
+                    "leaf_id": leaf_id,
+                    "mix_name": mix_name,
+                    "category_name": category["category_name"],
+                    "original_uint32_values": category_original,
+                    "source_percent_of_parent": category_source_percent,
+                    "target_uint32_values": category_target,
+                    "target_percent_of_parent": category_target_percent,
+                    "sampling_ratio": ""
+                    if category_ratio is None
+                    else f"{category_ratio:.12g}",
+                    "sampling_rate": category_sampling_rate,
+                }
+            )
+            path_definitions = sorted(
+                paths_by_leaf[leaf_id], key=lambda row: row["path_id"]
+            )
+            path_objects: list[dict[str, int]] = []
+            for path in path_definitions:
+                path_objects.append(
+                    {
+                        row["npy_uri"]: int(row["estimated_uint32_values"])
+                        for row in rows_by_path[path["path_id"]]
+                    }
+                )
+            path_originals = [sum(objects.values()) for objects in path_objects]
+            path_targets = _apportion_by_size(category_target, path_originals)
+            path_rows_html = _inventory_path_details(
+                path_definitions,
+                path_objects,
+                path_originals,
+                path_targets,
+                leaf_id=leaf_id,
+                mix_name=mix_name,
+                category_name=category["category_name"],
+                category_original=category_original,
+                category_target=category_target,
+                path_comparisons=path_comparisons,
+            )
+            category_sections.append(
+                _inventory_category_section(
+                    category["category_name"],
+                    source=category_original,
+                    source_percent=category_source_percent,
+                    target=category_target,
+                    target_percent=category_target_percent,
+                    sampling_rate=category_sampling_rate,
+                    sampling_class=category_class,
+                    path_rows_html=path_rows_html,
+                )
+            )
+        subcategory_detail_by_mix[mix_name] = _subcategory_detail_section(
+            mix_row,
+            _source_to_target_total_html(mix_row, mix_original, mix_target),
+            category_sections,
+        )
+
+    detail_sections = _family_detail_sections(
+        family_rows,
+        subcategory_detail_by_mix,
+        lambda row: _source_to_target_total_html(
+            row,
+            int(row["available_uint32_values"]),
+            int(row["target_uint32_values"]),
+        ),
+    )
+
+    _write_csv(
+        plot_data / "sampling-by-category.csv",
+        category_comparisons,
+        [
+            "leaf_id",
+            "mix_name",
+            "category_name",
+            "original_uint32_values",
+            "source_percent_of_parent",
+            "target_uint32_values",
+            "target_percent_of_parent",
+            "sampling_ratio",
+            "sampling_rate",
+        ],
+    )
+    _write_csv(
+        plot_data / "sampling-by-lower-group.csv",
+        path_comparisons,
+        [
+            "path_id",
+            "leaf_id",
+            "mix_name",
+            "category_name",
+            "lower_group",
+            "yaml_path",
+            "original_uint32_values",
+            "source_percent_of_parent",
+            "implied_target_uint32_values",
+            "implied_target_percent_of_parent",
+            "sampling_ratio",
+            "sampling_rate",
+            "unique_npy_count",
+        ],
+    )
+    aggregate_rate, _ = _sampling_rate_label(original_total, target_total)
+    sampling_audit_html = _sampling_audit_html(sampling_rate_rows)
+    report_html = (
+        '<!doctype html><html><head><meta charset="utf-8"><title>Dolma 3.5 sampling plan</title>'
+        + _interactive_report_style()
+        + "</head><body><h1>Dolma 3.5 Source Inventory and Sampling Plan</h1>"
+        + _summary_metrics(
+            [
+                ("Source tokens", _human_token_count(original_total)),
+                ("Target tokens", _human_token_count(target_total)),
+                ("Overall sampling", aggregate_rate),
+            ]
+        )
+        + sampling_audit_html
+        + f'<div class="mix-chart">{chart_rows}</div>'
+        + detail_sections
+        + _interactive_report_script()
+        + "</body></html>\n"
+    )
+    _write_text(phase / "report.html", report_html)
+    details = _build_inventory_details(
+        normalized_mix=normalized_mix,
+        normalized_paths=normalized_paths,
+        required_rows=required_rows,
+    )
+    _write_json(phase / "inventory-details.json", details)
+    return {
+        "source_family_count": details["source_family_count"],
+        "subcategory_count": details["subcategory_count"],
+        "source_count": details["source_count"],
+        "category_count": details["category_count"],
+        "lower_group_count": details["lower_group_count"],
+        "details_artifact": "inventory-details.json",
+        "report_artifact": "../report.html",
+    }
+
+
+def _render_report(
+    build: Path,
+    phase: Path,
+    allocations: Sequence[dict[str, Any]],
+    object_uses: Sequence[dict[str, Any]],
+    inventory: Sequence[dict[str, str]],
+    execution_units: Sequence[dict[str, Any]],
+    category_execution: Sequence[dict[str, Any]],
+) -> dict[str, int]:
+    plots = phase / "plots"
+    plot_data = phase / "plot-data"
+    plot_data.mkdir(exist_ok=False)
+    normalized_mix = _read_csv(build / "01-plan/resolution/normalized-mix.csv")
+    largest_units = sorted(
+        execution_units,
+        key=lambda row: int(row["estimated_peak_local_bytes"]),
+        reverse=True,
+    )[:40]
+    _write_csv(
+        plot_data / "largest-execution-units.csv",
+        largest_units,
+        [
+            "unit_id",
+            "leaf_id",
+            "mix_name",
+            "category_name",
+            "planned_uint32_values",
+            "input_npy_bytes",
+            "input_metadata_bytes",
+            "output_npy_bytes",
+            "estimated_output_metadata_bytes",
+            "estimated_selection_index_bytes",
+            "estimated_peak_local_bytes",
+            "max_unit_working_bytes",
+            "working_budget_utilization",
+        ],
+    )
+    _write_text(
+        plots / "largest-execution-units.svg",
+        _svg_bar_chart(
+            "Largest estimated execution-unit working sets",
+            [row["unit_id"] for row in largest_units],
+            [int(row["estimated_peak_local_bytes"]) / 1e9 for row in largest_units],
+            "GB",
+        ),
+    )
+    by_mix: dict[str, dict[str, Any]] = defaultdict(lambda: {"target": 0, "planned": 0})
+    for row in allocations:
+        entry = by_mix[row["mix_name"]]
+        entry["target"] += int(row["target_uint32_values"])
+        entry["planned"] += int(row["planned_uint32_values"])
+    top_targets = sorted(
+        by_mix.items(), key=lambda item: item[1]["target"], reverse=True
+    )[:40]
+    target_rows = [
+        {
+            "mix_name": name,
+            "target_uint32_values": values["target"],
+            "planned_uint32_values": values["planned"],
+        }
+        for name, values in top_targets
+    ]
+    target_total = sum(values["target"] for values in by_mix.values())
+    for row in target_rows:
+        row["target_percent"] = (
+            f"{100 * int(row['target_uint32_values']) / target_total:.8f}"
+            if target_total
+            else "0"
+        )
+    _write_csv(
+        plot_data / "target-by-mix.csv",
+        target_rows,
+        [
+            "mix_name",
+            "target_uint32_values",
+            "planned_uint32_values",
+            "target_percent",
+        ],
+    )
+    _write_text(
+        plots / "target-mix.svg",
+        _svg_bar_chart(
+            "Largest target mix entries by share of target",
+            [name for name, _ in top_targets],
+            [float(row["target_percent"]) for row in target_rows],
+            "% of target",
+            value_labels=[
+                _mix_plot_value_label(int(row["target_uint32_values"]), target_total)
+                for row in target_rows
+            ],
+            summary=(
+                f"Total target: {_human_token_count(target_total)} tokens "
+                f"({target_total:,})"
+            ),
+        ),
+    )
+    _write_csv(
+        plot_data / "proposal-target-residuals.csv",
+        allocations,
+        [
+            "leaf_id",
+            "mix_name",
+            "category_name",
+            "target_uint32_values",
+            "planned_uint32_values",
+            "target_residual_uint32_values",
+            "target_residual_bps_of_total",
+            "target_residual_fraction",
+        ],
+    )
+    normalized_paths = _read_csv(build / "01-plan/resolution/normalized-paths.csv")
+    allocation_by_leaf = {row["leaf_id"]: row for row in allocations}
+    inventory_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    inventory_by_path: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in inventory:
+        inventory_by_leaf[row["leaf_id"]].append(row)
+        inventory_by_path[row["path_id"]].append(row)
+    uses_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    uses_by_path: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in object_uses:
+        uses_by_leaf[row["leaf_id"]].append(row)
+        uses_by_path[row["path_id"]].append(row)
+    categories_by_mix: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    paths_by_leaf: dict[str, list[dict[str, Any]]] = defaultdict(list)
+    for row in normalized_mix:
+        categories_by_mix[row["mix_name"]].append(row)
+    for row in normalized_paths:
+        paths_by_leaf[row["leaf_id"]].append(row)
+
+    source_totals: dict[str, dict[str, int]] = defaultdict(
+        lambda: {"original": 0, "target": 0, "planned": 0}
+    )
+    original_by_leaf: dict[str, int] = {}
+    for category in normalized_mix:
+        leaf_id = category["leaf_id"]
+        original_objects = {
+            row["npy_uri"]: int(row["estimated_uint32_values"])
+            for row in inventory_by_leaf[leaf_id]
+        }
+        original = sum(original_objects.values())
+        planned = sum(
+            int(row["planned_uint32_values"]) for row in uses_by_leaf[leaf_id]
+        )
+        original_by_leaf[leaf_id] = original
+        entry = source_totals[category["mix_name"]]
+        entry["original"] += original
+        entry["target"] += int(category["target_uint32_values"])
+        entry["planned"] += planned
+
+    original_total = sum(entry["original"] for entry in source_totals.values())
+    proposed_total = sum(entry["planned"] for entry in source_totals.values())
+    # The per-mix walk below feeds only the two proposed-sampling CSVs, so it
+    # needs the mix names in descending proposed-output order. The report HTML
+    # comes from _render_execution_proposal_html.
+    ordered_mix_names = [
+        mix_name
+        for mix_name, _ in sorted(
+            source_totals.items(),
+            key=lambda item: (item[1]["planned"], item[1]["target"], item[0]),
+            reverse=True,
+        )
+    ]
+
+    category_sampling_rows: list[dict[str, Any]] = []
+    path_sampling_rows: list[dict[str, Any]] = []
+    for mix_name in ordered_mix_names:
+        for category in sorted(
+            categories_by_mix[mix_name], key=lambda row: int(row["category_index"])
+        ):
+            leaf_id = category["leaf_id"]
+            original = original_by_leaf[leaf_id]
+            target = int(category["target_uint32_values"])
+            planned = sum(
+                int(row["planned_uint32_values"]) for row in uses_by_leaf[leaf_id]
+            )
+            allocation = allocation_by_leaf.get(leaf_id)
+            if allocation is None:
+                minimum_repetition = 0
+                maximum_repetition = 0
+                total_object_uses = 0
+                unique_object_count = len(
+                    {row["npy_uri"] for row in inventory_by_leaf[leaf_id]}
+                )
+                selected_object_count = 0
+                dropped_object_count = unique_object_count
+                repeated_object_count = 0
+                partial_object_count = 0
+            else:
+                minimum_repetition = int(allocation["minimum_repetition"])
+                maximum_repetition = int(allocation["maximum_repetition"])
+                total_object_uses = int(allocation["total_object_uses"])
+                unique_object_count = int(allocation["unique_object_count"])
+                selected_object_count = int(allocation["selected_object_count"])
+                dropped_object_count = int(allocation["dropped_object_count"])
+                repeated_object_count = int(allocation["repeated_object_count"])
+                partial_object_count = int(allocation["partial_object_count"])
+            effective = planned / original if original else 0.0
+            category_sampling_rows.append(
+                {
+                    "leaf_id": leaf_id,
+                    "mix_name": mix_name,
+                    "category_name": category["category_name"],
+                    "original_uint32_values": original,
+                    "target_uint32_values": target,
+                    "proposed_uint32_values": planned,
+                    "token_change_from_original": planned - original,
+                    "effective_repetitions": f"{effective:.12g}",
+                    "minimum_repetition": minimum_repetition,
+                    "maximum_repetition": maximum_repetition,
+                    "total_object_uses": total_object_uses,
+                    "unique_object_count": unique_object_count,
+                    "selected_object_count": selected_object_count,
+                    "dropped_object_count": dropped_object_count,
+                    "repeated_object_count": repeated_object_count,
+                    "partial_object_count": partial_object_count,
+                }
+            )
+            for path in sorted(paths_by_leaf[leaf_id], key=lambda row: row["path_id"]):
+                path_original_objects = {
+                    row["npy_uri"]: int(row["estimated_uint32_values"])
+                    for row in inventory_by_path[path["path_id"]]
+                }
+                path_original = sum(path_original_objects.values())
+                path_uses = uses_by_path[path["path_id"]]
+                path_planned = sum(
+                    int(row["planned_uint32_values"]) for row in path_uses
+                )
+                path_repetitions = [int(row["repeat_count"]) for row in path_uses]
+                path_partials = [
+                    int(row.get("partial_target_uint32_values", 0)) for row in path_uses
+                ]
+                if not path_repetitions:
+                    path_repetitions = [0 for _ in path_original_objects]
+                    path_partials = [0 for _ in path_original_objects]
+                path_minimum = min(path_repetitions, default=0)
+                path_maximum = max(path_repetitions, default=0)
+                path_partial_count = sum(value > 0 for value in path_partials)
+                path_total_uses = sum(path_repetitions) + path_partial_count
+                path_selected = sum(
+                    repeat > 0 or partial > 0
+                    for repeat, partial in zip(path_repetitions, path_partials)
+                )
+                path_dropped = sum(
+                    repeat == 0 and partial == 0
+                    for repeat, partial in zip(path_repetitions, path_partials)
+                )
+                path_repeated = sum(
+                    int(row["planned_uint32_values"])
+                    > int(row["estimated_uint32_values"])
+                    for row in path_uses
+                )
+                path_effective = path_planned / path_original if path_original else 0.0
+                path_sampling_rows.append(
+                    {
+                        "path_id": path["path_id"],
+                        "leaf_id": leaf_id,
+                        "mix_name": mix_name,
+                        "category_name": category["category_name"],
+                        "lower_group": _path_subgroup(path["yaml_path"]),
+                        "yaml_path": path["yaml_path"],
+                        "original_uint32_values": path_original,
+                        "proposed_uint32_values": path_planned,
+                        "token_change_from_original": path_planned - path_original,
+                        "effective_repetitions": f"{path_effective:.12g}",
+                        "minimum_repetition": path_minimum,
+                        "maximum_repetition": path_maximum,
+                        "total_object_uses": path_total_uses,
+                        "unique_object_count": len(path_original_objects),
+                        "selected_object_count": path_selected,
+                        "dropped_object_count": path_dropped,
+                        "repeated_object_count": path_repeated,
+                        "partial_object_count": path_partial_count,
+                    }
+                )
+
+    _write_csv(
+        plot_data / "proposed-sampling-by-category.csv",
+        category_sampling_rows,
+        list(category_sampling_rows[0]) if category_sampling_rows else [],
+    )
+    _write_csv(
+        plot_data / "proposed-sampling-by-lower-group.csv",
+        path_sampling_rows,
+        list(path_sampling_rows[0]) if path_sampling_rows else [],
+    )
+    report = _render_execution_proposal_html(
+        execution_units=execution_units,
+        category_execution=category_execution,
+        planned_total=proposed_total,
+    )
+    _write_text(phase / "report.html", report)
+    return {
+        "source_uint32_values": original_total,
+        "proposed_uint32_values": proposed_total,
+        "target_uint32_values": target_total,
+    }
+
+
+def validate_build(args: argparse.Namespace) -> None:
+    """Re-check every phase summary a build has recorded.
+
+    Writes no artifacts. Prints one JSON check report to stdout and raises
+    PreparationError if any check failed.
+    """
+
+    build = args.build.resolve()
+    manifest = _load_build(build)
+    checks: list[tuple[str, bool, str]] = []
+    plan_failures = _read_csv(build / "01-plan/resolution/resolution-failures.csv")
+    checks.append(("plan_failures", not plan_failures, str(len(plan_failures))))
+    inventory_summary_path = build / "01-plan/inventory/inventory-summary.json"
+    checks.append(
+        (
+            "inventory_exists",
+            inventory_summary_path.is_file(),
+            str(inventory_summary_path),
+        )
+    )
+    if inventory_summary_path.is_file():
+        with inventory_summary_path.open() as f:
+            summary = json.load(f)
+        for name in (
+            "missing_objects",
+            "path_resolution_failures",
+            "invalid_npy_sizes",
+            "head_errors",
+            "sampling_rate_limit_failures",
+        ):
+            value = summary.get(
+                name,
+                summary.get("direct_resolution_failures", 0)
+                if name == "path_resolution_failures"
+                else 0,
+            )
+            checks.append((name, int(value) == 0, str(value)))
+    validation_path = build / "01-plan/execution/validation-summary.json"
+    checks.append(("execution_exists", validation_path.is_file(), str(validation_path)))
+    if validation_path.is_file():
+        with validation_path.open() as f:
+            proposal = json.load(f)
+        checks.append(("execution_passed", bool(proposal["passed"]), str(proposal)))
+        try:
+            layout = _validate_execution_layout(build)
+            checks.append(("execution_layout_current", True, str(layout)))
+        except PreparationError as exc:
+            checks.append(("execution_layout_current", False, str(exc)))
+    combined_report_path = build / "01-plan/report.html"
+    checks.append(
+        (
+            "combined_plan_report_exists",
+            combined_report_path.is_file(),
+            str(combined_report_path),
+        )
+    )
+    checks.append(
+        (
+            "stage_reports_consolidated",
+            not (build / "01-plan/resolution/report.html").exists()
+            and not (build / "01-plan/inventory/report.html").exists()
+            and not (build / "01-plan/execution/report.html").exists(),
+            "individual stage reports must not remain after composition",
+        )
+    )
+    preflight_path = build / "02-preflight/preflight-summary.json"
+    if preflight_path.is_file():
+        with preflight_path.open() as f:
+            preflight = json.load(f)
+        checks.append(("preflight_passed", bool(preflight["passed"]), str(preflight)))
+    output_path = build / "03-output-validation/output-summary.json"
+    if output_path.is_file():
+        with output_path.open() as f:
+            output = json.load(f)
+        checks.append(("output_validation_passed", bool(output["passed"]), str(output)))
+    failed = [check for check in checks if not check[1]]
+    print(
+        json.dumps(
+            {"build_id": manifest["build_id"], "checks": checks, "passed": not failed},
+            indent=2,
+        )
+    )
+    if failed:
+        raise PreparationError(f"Build has {len(failed)} failed validation check(s)")
diff --git a/scripts/resharding/__init__.py b/scripts/resharding/__init__.py
new file mode 100644
index 00000000..c26637dc
--- /dev/null
+++ b/scripts/resharding/__init__.py
@@ -0,0 +1,6 @@
+"""Mixture-agnostic resharding helpers.
+
+`dispatch` builds poormanray command lines for the worker lifecycle. The
+planning and materialization workflow for the Dolma 3.5 14T mix is in
+`scripts/dolma3p5_resharding/` and consumes these builders.
+"""
diff --git a/scripts/resharding/dispatch.py b/scripts/resharding/dispatch.py
new file mode 100644
index 00000000..5c768bd8
--- /dev/null
+++ b/scripts/resharding/dispatch.py
@@ -0,0 +1,227 @@
+"""Build poormanray commands for a complete resharding worker lifecycle."""
+
+from __future__ import annotations
+
+from collections.abc import Sequence
+from pathlib import Path
+
+
+def _base_command(
+    action: str,
+    *,
+    cluster: str,
+    region: str | None,
+    project: str | None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    if not cluster.strip():
+        raise ValueError("cluster must not be empty")
+    command = [*(runner or ("pmr",)), action, "--name", cluster]
+    if project:
+        command.extend(("--project", project))
+    if region:
+        command.extend(("--region", region))
+    return command
+
+
+def _add_instance_options(
+    command: list[str],
+    *,
+    instance_ids: Sequence[str] = (),
+    parallelism: int | None = None,
+    ssh_key_path: str | Path | None = None,
+) -> list[str]:
+    for instance_id in instance_ids:
+        if not instance_id.strip():
+            raise ValueError("instance IDs must not be empty")
+        command.extend(("--instance-id", instance_id))
+    if parallelism is not None:
+        if parallelism <= 0:
+            raise ValueError("parallelism must be positive")
+        command.extend(("--parallelism", str(parallelism)))
+    if ssh_key_path is not None:
+        command.extend(("--ssh-key-path", str(ssh_key_path)))
+    return command
+
+
+def build_poormanray_create_command(
+    *,
+    cluster: str,
+    project: str,
+    region: str,
+    number: int,
+    instance_type: str,
+    storage_type: str,
+    storage_size_gib: int,
+    parallelism: int | None = None,
+    detach: bool = False,
+    ssh_key_path: str | Path | None = None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    """Build the command that creates missing workers for a cluster."""
+
+    if number <= 0:
+        raise ValueError("number must be positive")
+    if storage_size_gib <= 0:
+        raise ValueError("storage_size_gib must be positive")
+    command = _base_command(
+        "create", cluster=cluster, project=project, region=region, runner=runner
+    )
+    command.extend(
+        (
+            "--number",
+            str(number),
+            "--instance-type",
+            instance_type,
+            "--storage-type",
+            storage_type,
+            "--storage-size",
+            str(storage_size_gib),
+        )
+    )
+    if detach:
+        command.append("--detach")
+    return _add_instance_options(
+        command,
+        parallelism=parallelism,
+        ssh_key_path=ssh_key_path,
+    )
+
+
+def build_poormanray_instance_command(
+    action: str,
+    *,
+    cluster: str,
+    project: str,
+    region: str,
+    instance_ids: Sequence[str],
+    parallelism: int | None = None,
+    detach: bool = False,
+    ssh_key_path: str | Path | None = None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    """Build a wait, resume, or pause command scoped to explicit workers."""
+
+    if action not in {"wait", "resume", "pause"}:
+        raise ValueError(f"unsupported instance action: {action}")
+    command = _base_command(
+        action, cluster=cluster, project=project, region=region, runner=runner
+    )
+    if detach and action != "wait":
+        command.append("--detach")
+    return _add_instance_options(
+        command,
+        instance_ids=instance_ids,
+        parallelism=parallelism if action != "wait" else None,
+        ssh_key_path=ssh_key_path if action == "wait" else None,
+    )
+
+
+def build_poormanray_transfer_command(
+    *,
+    cluster: str,
+    project: str,
+    region: str,
+    transfers: Sequence[tuple[str | Path, str]],
+    instance_ids: Sequence[str] = (),
+    parallelism: int | None = None,
+    ssh_key_path: str | Path | None = None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    """Build a transfer command for worker bootstrap files."""
+
+    if not transfers:
+        raise ValueError("at least one transfer is required")
+    command = _base_command(
+        "transfer", cluster=cluster, project=project, region=region, runner=runner
+    )
+    for source, destination in transfers:
+        command.extend(("--source", f"{source}:{destination}"))
+    return _add_instance_options(
+        command,
+        instance_ids=instance_ids,
+        parallelism=parallelism,
+        ssh_key_path=ssh_key_path,
+    )
+
+
+def build_poormanray_run_command(
+    *,
+    cluster: str,
+    project: str,
+    region: str,
+    remote_command: str,
+    instance_ids: Sequence[str] = (),
+    parallelism: int | None = None,
+    ssh_key_path: str | Path | None = None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    """Build a synchronous command to run on selected workers."""
+
+    if not remote_command.strip():
+        raise ValueError("remote_command must not be empty")
+    command = _base_command(
+        "run", cluster=cluster, project=project, region=region, runner=runner
+    )
+    command.extend(("--command", remote_command))
+    return _add_instance_options(
+        command,
+        instance_ids=instance_ids,
+        parallelism=parallelism,
+        ssh_key_path=ssh_key_path,
+    )
+
+
+def build_poormanray_setup_dolma_command(
+    *,
+    cluster: str,
+    project: str,
+    region: str,
+    instance_ids: Sequence[str] = (),
+    parallelism: int | None = None,
+    ssh_key_path: str | Path | None = None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    """Build the command that installs Dolma and s5cmd on selected workers."""
+
+    command = _base_command(
+        "setup-dolma-python",
+        cluster=cluster,
+        project=project,
+        region=region,
+        runner=runner,
+    )
+    return _add_instance_options(
+        command,
+        instance_ids=instance_ids,
+        parallelism=parallelism,
+        ssh_key_path=ssh_key_path,
+    )
+
+
+def build_poormanray_map_command(
+    *,
+    cluster: str,
+    script_dir: str | Path,
+    region: str | None = None,
+    project: str | None = None,
+    spindown: bool,
+    instance_ids: Sequence[str] = (),
+    parallelism: int | None = None,
+    ssh_key_path: str | Path | None = None,
+    runner: Sequence[str] | None = None,
+) -> list[str]:
+    """Build a poormanray command that maps scripts over selected workers."""
+
+    command = _base_command(
+        "map", cluster=cluster, project=project, region=region, runner=runner
+    )
+    command.extend(("--script", str(script_dir)))
+    if spindown:
+        command.append("--spindown")
+    return _add_instance_options(
+        command,
+        instance_ids=instance_ids,
+        parallelism=parallelism,
+        ssh_key_path=ssh_key_path,
+    )
diff --git a/tests/python/test_dolma3p5_resharding.py b/tests/python/test_dolma3p5_resharding.py
new file mode 100644
index 00000000..a36761be
--- /dev/null
+++ b/tests/python/test_dolma3p5_resharding.py
@@ -0,0 +1,3378 @@
+import argparse
+import csv
+import gzip
+import io
+import json
+import os
+import shlex
+import subprocess
+import sys
+import tempfile
+import time
+import unittest
+from contextlib import redirect_stdout
+from pathlib import Path
+from threading import Event, Lock
+from types import SimpleNamespace
+from unittest.mock import ANY, MagicMock, call, patch
+from xml.etree import ElementTree
+
+import numpy as np
+import yaml
+from rich.console import Console
+
+from dolma.tokenizer.document_selection import (
+    DOCUMENT_SELECTION_ALGORITHM,
+    create_document_selection,
+)
+from dolma.tokenizer.reshard import (
+    RESHARDING_MANIFEST_SCHEMA_VERSION,
+    ReshardingConfig,
+    ReshardingManifestConfig,
+)
+from dolma.tokenizer.reshard import _run_s5cmd as run_s5cmd
+from dolma.tokenizer.reshard import (
+    destination_has_objects,
+    merge_group,
+    reshard,
+    upload_to_s3,
+)
+from scripts.dolma3p5_resharding.materialize import (
+    ClusterInstance,
+    MaterializationGroup,
+    PlannedWorkerGroup,
+    WorkerAssignment,
+    _execute_materialization_groups,
+    _map_command,
+    _pause_workers_after_failure,
+    _planned_worker_groups,
+    _prepare_workers,
+    _print_dispatch,
+    _provision_batches,
+    _resume_worker_groups_in_batches,
+    _resume_workers_in_batches,
+    _retag_cluster_instances,
+    _run_compact_process,
+    _run_selected_preflight,
+    _safe_path_launcher_payload,
+    _select_units,
+    _stage_worker_assignments,
+    _status_detail,
+    _verify_materialized_units,
+    _wait_command,
+    _wait_for_workers_to_stop,
+    _worker_counts_for_groups,
+    _worker_log_command,
+    _worker_log_message,
+    _worker_log_parts,
+    _worker_log_snapshots,
+)
+from scripts.dolma3p5_resharding.materialize import (
+    build_parser as build_materialize_parser,
+)
+from scripts.dolma3p5_resharding.workflow import (
+    DEFAULT_REGION,
+)
+from scripts.dolma3p5_resharding.workflow import (
+    DOCUMENT_SELECTION_ALGORITHM as WORKFLOW_DOCUMENT_SELECTION_ALGORITHM,
+)
+from scripts.dolma3p5_resharding.workflow import (
+    EXECUTION_UNIT_INDEX_WIDTH,
+    EXPECTED_MANIFEST_SCHEMA_VERSION,
+    PreparationError,
+    S3Object,
+    _allocate_object_sampling,
+    _category_output_directory,
+    _execution_unit_work,
+    _filter_execution_units,
+    _finalize_inventory,
+    _load_catalog,
+    _parse_s5cmd_jsonl,
+    _partition_object_uses,
+    _planned_output_shards,
+    _select_worker_instance,
+    _self_contained_launcher,
+    _source_relative_directory,
+    _unit_selection_digest,
+    _validate_execution_layout,
+    collect_inventory,
+    normalize_region,
+    plan_build,
+    preflight_build,
+    propose_configs,
+    refresh_inventory_details,
+    validate_build,
+    verify_output,
+)
+
+WORKER_STORAGE_SCRIPT = Path(__file__).resolve().parents[2] / "scripts/dolma3p5_resharding/setup_worker_storage.sh"
+
+
+class TestDolma35ReshardingPreparation(unittest.TestCase):
+    def test_output_shard_planning_is_size_based_and_bounded(self):
+        target = 64 * 1024**3
+        self.assertEqual(
+            _planned_output_shards(
+                output_npy_bytes=target // 2,
+                input_view_count=20,
+                target_output_shard_bytes=target,
+                max_output_shards_per_unit=8,
+            ),
+            1,
+        )
+        self.assertEqual(
+            _planned_output_shards(
+                output_npy_bytes=target * 3,
+                input_view_count=20,
+                target_output_shard_bytes=target,
+                max_output_shards_per_unit=8,
+            ),
+            3,
+        )
+        self.assertEqual(
+            _planned_output_shards(
+                output_npy_bytes=target * 20,
+                input_view_count=20,
+                target_output_shard_bytes=target,
+                max_output_shards_per_unit=8,
+            ),
+            8,
+        )
+        self.assertEqual(
+            _planned_output_shards(
+                output_npy_bytes=target * 3,
+                input_view_count=2,
+                target_output_shard_bytes=target,
+                max_output_shards_per_unit=8,
+            ),
+            2,
+        )
+
+    def test_execution_work_and_worker_grid_account_for_cpu_and_disk(self):
+        rows = [
+            {
+                "estimated_uint32_values": 100,
+                "repeat_count": 1,
+                "partial_target_uint32_values": 30,
+            }
+        ]
+        work = _execution_unit_work(rows, document_selection_work_passes=2)
+        self.assertEqual(work["planned_uint32_values"], 130)
+        self.assertEqual(work["document_selection_source_uint32_values"], 100)
+        self.assertEqual(work["estimated_work_uint32_values"], 330)
+
+        grid = [
+            {
+                "instance_type": "i4i.2xlarge",
+                "vcpus": 8,
+                "local_nvme_devices": 1,
+                "local_nvme_bytes": 1_875,
+                "max_estimated_work_uint32_values": 150,
+            },
+            {
+                "instance_type": "i4i.4xlarge",
+                "vcpus": 16,
+                "local_nvme_devices": 1,
+                "local_nvme_bytes": 3_750,
+                "max_estimated_work_uint32_values": 300,
+            },
+            {
+                "instance_type": "i4i.8xlarge",
+                "vcpus": 32,
+                "local_nvme_devices": 2,
+                "local_nvme_bytes": 7_500,
+                "max_estimated_work_uint32_values": None,
+            },
+        ]
+        self.assertEqual(
+            _select_worker_instance(
+                grid,
+                estimated_peak_local_bytes=1_000,
+                estimated_work_uint32_values=100,
+                disk_headroom_ratio=1.1,
+            )["instance_type"],
+            "i4i.2xlarge",
+        )
+        self.assertEqual(
+            _select_worker_instance(
+                grid,
+                estimated_peak_local_bytes=2_000,
+                estimated_work_uint32_values=100,
+                disk_headroom_ratio=1.1,
+            )["instance_type"],
+            "i4i.4xlarge",
+        )
+        largest = _select_worker_instance(
+            grid,
+            estimated_peak_local_bytes=1_000,
+            estimated_work_uint32_values=330,
+            disk_headroom_ratio=1.1,
+        )
+        self.assertEqual(largest["instance_type"], "i4i.8xlarge")
+        self.assertEqual(largest["storage_layout"], "raid0")
+
+    def test_materialize_dry_run_uses_terse_cli_output(self):
+        output = io.StringIO()
+        with redirect_stdout(output):
+            _print_dispatch(
+                "category-example",
+                [
+                    {
+                        "leaf_id": "000:example",
+                        "unit_id": "00000000",
+                        "planned_uint32_values": "1000000000",
+                        "planned_output_shard_count": "1",
+                        "estimated_peak_local_bytes": "2000000000",
+                        "destination_prefix": "s3://bucket/output/00000000",
+                    }
+                ],
+                Path("/tmp/launchers"),
+                [("create missing workers", ["pmr", "create", "--number", "1"])],
+                1,
+                False,
+            )
+        rendered = output.getvalue()
+        self.assertIn("Dry run", rendered)
+        self.assertIn("Selection", rendered)
+        self.assertIn("example", rendered)
+        self.assertIn("1 category · 1 unit · 1 worker", rendered)
+        self.assertIn("commands:\npmr create --number 1", rendered)
+        self.assertNotIn("Worker lifecycle", rendered)
+        self.assertNotIn("create missing workers:", rendered)
+
+    def test_materialize_execute_summary_omits_dry_run_details(self):
+        output = io.StringIO()
+        with redirect_stdout(output):
+            _print_dispatch(
+                "category-example",
+                [
+                    {
+                        "leaf_id": "000:example",
+                        "unit_id": "00000000",
+                        "planned_uint32_values": "1000000000",
+                        "planned_output_shard_count": "1",
+                        "estimated_peak_local_bytes": "2000000000",
+                        "destination_prefix": "s3://bucket/output/00000000",
+                    }
+                ],
+                Path("/tmp/launchers"),
+                [("create missing workers", ["pmr", "create", "--number", "1"])],
+                1,
+                True,
+                cluster="dolma3p5-14t",
+                project="oe-other",
+                region="us-east-1",
+            )
+        rendered = output.getvalue()
+        self.assertIn("Execution", rendered)
+        self.assertIn("Selection", rendered)
+        self.assertIn("1 category · 1 unit · 1 worker", rendered)
+        self.assertIn("Cluster", rendered)
+        self.assertIn("dolma3p5-14t", rendered)
+        self.assertIn("Project", rendered)
+        self.assertIn("oe-other", rendered)
+        self.assertIn("Region", rendered)
+        self.assertIn("us-east-1", rendered)
+        self.assertNotIn("category-example", rendered)
+        self.assertNotIn("commands:", rendered)
+
+    def test_materialize_summary_pluralizes_categories(self):
+        output = io.StringIO()
+        rows = [
+            {
+                "leaf_id": f"{index:03d}:example",
+                "unit_id": f"{index:08d}",
+                "planned_uint32_values": "1000000000",
+                "planned_output_shard_count": "1",
+                "estimated_peak_local_bytes": "2000000000",
+                "destination_prefix": f"s3://bucket/output/{index:08d}",
+            }
+            for index in range(2)
+        ]
+        with redirect_stdout(output):
+            _print_dispatch(
+                "all",
+                rows,
+                Path("/tmp/launchers"),
+                [],
+                2,
+                True,
+            )
+
+        rendered = output.getvalue()
+        self.assertIn("2 categories · 2 units · 2 workers", rendered)
+        self.assertNotIn("categorys", rendered)
+        self.assertNotIn("pmr create", rendered)
+        self.assertNotIn("units:", rendered)
+
+    def test_compact_process_discards_success_output(self):
+        output = io.StringIO()
+        console = Console(file=output, force_terminal=False, color_system=None)
+        return_code = _run_compact_process(
+            "wait for workers",
+            [
+                sys.executable,
+                "-c",
+                (
+                    "print('Waiting for instances... 0/2 ready (0s)'); "
+                    "print('  · worker-0000 [running]'); "
+                    "print('Waiting for instances... 2/2 ready (10s)')"
+                ),
+            ],
+            console=console,
+        )
+        rendered = output.getvalue()
+        self.assertEqual(return_code, 0)
+        self.assertIn("wait for workers", rendered)
+        self.assertIn("✓", rendered)
+        self.assertNotIn("Waiting for instances", rendered)
+        self.assertNotIn("worker-0000", rendered)
+
+    def test_compact_process_streams_child_output_when_verbose(self):
+        output = io.StringIO()
+        return_code = _run_compact_process(
+            "prepare local NVMe",
+            [sys.executable, "-c", "print('worker setup detail')"],
+            console=Console(file=output, force_terminal=False, color_system=None),
+            verbose=True,
+        )
+        self.assertEqual(return_code, 0)
+        self.assertIn("worker setup detail", output.getvalue())
+
+    def test_compact_process_bounds_and_redacts_failure_output(self):
+        output = io.StringIO()
+        console = Console(file=output, force_terminal=False, color_system=None)
+        return_code = _run_compact_process(
+            "create workers",
+            [
+                sys.executable,
+                "-c",
+                (
+                    "import sys; "
+                    "[print(f'line-{index}') for index in range(20)]; "
+                    "print('aws_secret_access_key=do-not-print'); "
+                    "sys.exit(3)"
+                ),
+            ],
+            console=console,
+        )
+        rendered = output.getvalue()
+        self.assertEqual(return_code, 3)
+        self.assertIn("✗ create workers", rendered)
+        self.assertIn("last output:", rendered)
+        self.assertNotIn("line-0\n", rendered)
+        self.assertIn("line-19", rendered)
+        self.assertIn("aws_secret_access_key=<redacted>", rendered)
+        self.assertNotIn("do-not-print", rendered)
+
+    def test_compact_status_hides_poormanray_cluster_log_mislabeled_as_project(self):
+        self.assertIsNone(
+            _status_detail(
+                "prepare local NVMe",
+                "[INFO][21:31:31] Running command on instances with "
+                "project=dolma3p5-14t in region us-east-1 (aws)",
+            )
+        )
+
+    def test_worker_log_message_removes_envelope_but_preserves_errors(self):
+        self.assertEqual(
+            _worker_log_parts("[2026-08-05 15:38:11 main.dolma.__main__ INFO] merge 50.0%"),
+            ("2026-08-05 15:38:11", "merge 50.0%"),
+        )
+        self.assertEqual(
+            _worker_log_message("[2026-08-05 15:38:11 main.dolma.__main__ INFO] merge 50.0%"),
+            "merge 50.0%",
+        )
+        self.assertEqual(
+            _worker_log_message("[2026-08-05 15:38:11 main.dolma.__main__ ERROR] disk full"),
+            "error · disk full",
+        )
+
+    def test_materialize_waits_for_worker_spindown(self):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            profile=None,
+            completion_poll_seconds=30,
+            verbose=True,
+        )
+        describe = MagicMock(
+            side_effect=[
+                [
+                    ClusterInstance("i-first", "running", "i4i.2xlarge", "oe-other"),
+                    ClusterInstance("i-second", "stopping", "i4i.2xlarge", "oe-other"),
+                ],
+                [
+                    ClusterInstance("i-first", "running", "i4i.2xlarge", "oe-other"),
+                    ClusterInstance("i-second", "running", "i4i.2xlarge", "oe-other"),
+                ],
+                [
+                    ClusterInstance("i-first", "stopped", "i4i.2xlarge", "oe-other"),
+                    ClusterInstance("i-second", "stopped", "i4i.2xlarge", "oe-other"),
+                ],
+            ]
+        )
+        sleep = MagicMock()
+        output = io.StringIO()
+        with patch(
+            "scripts.dolma3p5_resharding.materialize._worker_log_snapshots",
+            side_effect=[
+                {
+                    "i-first": {
+                        "statuses": {"00000000": "running"},
+                        "logs": {
+                            "00000000.log": ("Downloading exact manifest objects",),
+                        },
+                    }
+                },
+                {
+                    "i-first": {
+                        "statuses": {"00000000": "running"},
+                        "logs": {
+                            "00000000.log": (
+                                "Downloading exact manifest objects",
+                                (
+                                    "[2026-08-05 15:38:11 main.dolma.__main__ INFO] "
+                                    "merge 50.0% · 10B/20B tokens · 40M tokens/s"
+                                ),
+                            ),
+                        },
+                    },
+                    "i-second": {
+                        "statuses": {"00000001": "running"},
+                        "logs": {
+                            "00000001.log": ("Other worker progress",),
+                        },
+                    },
+                },
+            ],
+        ):
+            _wait_for_workers_to_stop(
+                args,
+                ["i-first", "i-second"],
+                2,
+                "test-run",
+                describe=describe,
+                sleep=sleep,
+                console=Console(file=output, force_terminal=False, color_system=None),
+            )
+        self.assertEqual(sleep.call_args_list, [call(30), call(30)])
+        rendered = output.getvalue()
+        self.assertIn("2 units · 1 running · 1 stopping · 0 stopped", rendered)
+        self.assertIn("] [-first] 00000000 · running", rendered)
+        self.assertIn("] [-first] unit 00000000", rendered)
+        self.assertEqual(rendered.count("unit 00000000"), 1)
+        self.assertIn("Downloading exact manifest objects", rendered)
+        self.assertEqual(rendered.count("Downloading exact manifest objects"), 1)
+        self.assertIn("merge 50.0% · 10B/20B tokens · 40M tokens/s", rendered)
+        self.assertNotIn("main.dolma.__main__ INFO", rendered)
+        self.assertIn("] [second] Other worker progress", rendered)
+        self.assertIn("[2026-08-05 15:38:11] [-first] merge 50.0%", rendered)
+        self.assertNotIn("[i-first]", rendered)
+        self.assertNotIn("[i-second]", rendered)
+        self.assertNotIn("worker 01", rendered)
+        self.assertIn("materialization workers stopped", rendered)
+
+    def test_materialize_status_tracks_workers_in_different_lifecycle_stages(self):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            profile=None,
+            completion_poll_seconds=30,
+            verbose=False,
+        )
+        describe = MagicMock(
+            side_effect=[
+                [
+                    ClusterInstance("i-ready", "running", "i4i.2xlarge", "oe-other"),
+                    ClusterInstance("i-slow", "running", "i4i.2xlarge", "oe-other"),
+                ],
+                [
+                    ClusterInstance("i-ready", "stopped", "i4i.2xlarge", "oe-other"),
+                    ClusterInstance("i-slow", "running", "i4i.2xlarge", "oe-other"),
+                ],
+                [
+                    ClusterInstance("i-ready", "stopped", "i4i.2xlarge", "oe-other"),
+                    ClusterInstance("i-slow", "stopped", "i4i.2xlarge", "oe-other"),
+                ],
+            ]
+        )
+        worker_stages = {"i-ready": "materializing", "i-slow": "waiting"}
+        stage_lock = Lock()
+        all_dispatched = Event()
+
+        sleep_calls = 0
+
+        def advance_slow_worker(_seconds):
+            nonlocal sleep_calls
+            sleep_calls += 1
+            if sleep_calls == 1:
+                with stage_lock:
+                    worker_stages["i-slow"] = "materializing"
+                all_dispatched.set()
+
+        sleep = MagicMock(side_effect=advance_slow_worker)
+        output = io.StringIO()
+
+        _wait_for_workers_to_stop(
+            args,
+            ["i-ready", "i-slow"],
+            2,
+            "test-run",
+            describe=describe,
+            sleep=sleep,
+            console=Console(file=output, force_terminal=False, color_system=None, width=300),
+            worker_stages=worker_stages,
+            stage_lock=stage_lock,
+            all_dispatched=all_dispatched,
+        )
+
+        rendered = output.getvalue()
+        self.assertIn("1 waiting", rendered)
+        self.assertIn("1 materializing", rendered)
+        self.assertIn("1 stopped", rendered)
+        self.assertIn("2 stopped", rendered)
+        self.assertIn("materialization workers stopped", rendered)
+
+    @patch("scripts.dolma3p5_resharding.materialize.subprocess.run")
+    def test_verbose_worker_logs_are_complete_for_every_worker(self, run):
+        run.return_value = SimpleNamespace(
+            returncode=0,
+            stdout=(
+                "Instance i-first:\n"
+                "stdout: @@DOLMA_STATUS@@\t00000000\trunning\n"
+                "@@DOLMA_LOG_BEGIN@@\t00000000.log\t0\t123\t123\n"
+                "first line\n"
+                "12.00%  12 GB / 100 GB (1.20 GB/s) 1m left (0/1)\r"
+                "24.00%  24 GB / 100 GB (1.18 GB/s) 1m left (0/1)\r"
+                "second line\n"
+                "@@DOLMA_LOG_END@@\t00000000.log\t123\tOK\n"
+                "stderr:\n\n"
+                "Instance i-second:\n"
+                "stdout: @@DOLMA_STATUS@@\t00000001\tsucceeded\n"
+                "@@DOLMA_LOG_BEGIN@@\t00000001.log\t0\t50\t50\n"
+                "other worker line\n"
+                "@@DOLMA_LOG_END@@\t00000001.log\t50\tOK\n"
+                "stderr:\n"
+            ),
+            stderr="",
+        )
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=2,
+            ssh_key_path=None,
+        )
+        snapshots = _worker_log_snapshots(
+            args,
+            ["i-first", "i-second"],
+            "test-run",
+        )
+        self.assertEqual(
+            snapshots["i-first"]["logs"]["00000000.log"],
+            (
+                "first line",
+                "12.00%  12 GB / 100 GB (1.20 GB/s) 1m left (0/1)",
+                "24.00%  24 GB / 100 GB (1.18 GB/s) 1m left (0/1)",
+                "second line",
+            ),
+        )
+        self.assertEqual(
+            snapshots["i-second"]["logs"]["00000001.log"],
+            ("other worker line",),
+        )
+        self.assertEqual(snapshots["i-first"]["offsets"], {"00000000.log": 123})
+        self.assertEqual(snapshots["i-second"]["offsets"], {"00000001.log": 50})
+
+    @patch("scripts.dolma3p5_resharding.materialize.subprocess.run")
+    def test_worker_log_capture_retries_an_unacknowledged_page(self, run):
+        run.return_value = SimpleNamespace(
+            returncode=0,
+            stdout=(
+                "Instance i-first:\n"
+                "stdout: @@DOLMA_STATUS@@\t00000000\trunning\n"
+                "@@DOLMA_LOG_BEGIN@@\t00000000.log\t100\t140\t200\n"
+                "partial progress record\n"
+                "@@DOLMA_LOG_END@@\t00000000.log\t140\n"
+                "stderr:\n"
+            ),
+            stderr="",
+        )
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=1,
+            ssh_key_path=None,
+        )
+        offsets = {("i-first", "00000000.log"): 100}
+
+        snapshots = _worker_log_snapshots(
+            args,
+            ["i-first"],
+            "test-run",
+            offsets,
+        )
+
+        self.assertEqual(snapshots["i-first"]["logs"], {})
+        self.assertEqual(snapshots["i-first"]["offsets"], {})
+        command = " ".join(run.call_args.args[0])
+        self.assertIn("00000000.log", command)
+        self.assertIn("100", command)
+
+    @patch("scripts.dolma3p5_resharding.materialize.subprocess.run")
+    def test_worker_log_capture_resumes_after_a_successful_page(self, run):
+        run.side_effect = [
+            SimpleNamespace(
+                returncode=0,
+                stdout=(
+                    "Instance i-first:\n"
+                    "stdout: @@DOLMA_STATUS@@\t00000000\trunning\n"
+                    "@@DOLMA_LOG_BEGIN@@\t00000000.log\t0\t100\t180\n"
+                    "first progress record\n"
+                    "@@DOLMA_LOG_END@@\t00000000.log\t100\tOK\n"
+                    "stderr:\n"
+                ),
+                stderr="",
+            ),
+            SimpleNamespace(
+                returncode=0,
+                stdout=(
+                    "Instance i-first:\n"
+                    "stdout: @@DOLMA_STATUS@@\t00000000\trunning\n"
+                    "@@DOLMA_LOG_BEGIN@@\t00000000.log\t100\t180\t180\n"
+                    "second progress record\n"
+                    "@@DOLMA_LOG_END@@\t00000000.log\t180\tOK\n"
+                    "stderr:\n"
+                ),
+                stderr="",
+            ),
+        ]
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=1,
+            ssh_key_path=None,
+        )
+
+        first = _worker_log_snapshots(args, ["i-first"], "test-run")
+        offsets = {("i-first", log_name): offset for log_name, offset in first["i-first"]["offsets"].items()}
+        second = _worker_log_snapshots(
+            args,
+            ["i-first"],
+            "test-run",
+            offsets,
+        )
+
+        self.assertEqual(
+            first["i-first"]["logs"]["00000000.log"],
+            ("first progress record",),
+        )
+        self.assertEqual(
+            second["i-first"]["logs"]["00000000.log"],
+            ("second progress record",),
+        )
+        self.assertEqual(second["i-first"]["offsets"], {"00000000.log": 180})
+        second_command = " ".join(run.call_args_list[1].args[0])
+        self.assertIn("00000000.log", second_command)
+        self.assertIn("100", second_command)
+
+    def test_materialize_verifies_selected_output_sizes_and_metadata(self):
+        args = SimpleNamespace(
+            profile=None,
+            region="us-east-1",
+            parallelism=2,
+        )
+        row = {
+            "unit_id": "00000000",
+            "destination_prefix": "s3://bucket/output/00000000",
+            "planned_uint32_values": "10",
+            "planned_output_shard_count": "1",
+            "allowed_materialized_target_residual_uint32_values": "0",
+        }
+        objects = [
+            S3Object("bucket", "output/00000000/000000.npy", 40),
+            S3Object("bucket", "output/00000000/000000.csv.gz", 20),
+        ]
+        output = io.StringIO()
+        with patch(
+            "scripts.dolma3p5_resharding.materialize._list_prefix",
+            return_value=objects,
+        ):
+            checks = _verify_materialized_units(
+                args,
+                [row],
+                client=MagicMock(),
+                console=Console(file=output, force_terminal=False, color_system=None),
+            )
+        self.assertEqual(checks[0].actual_uint32_values, 10)
+        self.assertEqual(checks[0].metadata_count, 1)
+        self.assertIn("materialization verified  1/1 units", output.getvalue())
+
+    def test_generated_materialize_launcher_uses_safe_import_directory(self):
+        launcher = _self_contained_launcher(
+            unit_id="00000000",
+            config_name="00000000.yaml",
+            config_text="destination_prefix: /tmp/output\n",
+            manifest_name="00000000.csv",
+            manifest_text="npy_uri,metadata_uri\n",
+        )
+        self.assertIn("export PYTHONSAFEPATH=1\ncd /tmp\n", launcher)
+        self.assertIn('"$python_bin" -P -m dolma.tokenizer.reshard', launcher)
+        with tempfile.TemporaryDirectory() as temp_dir:
+            path = Path(temp_dir) / "launcher.sh"
+            path.write_text(launcher)
+            self.assertEqual(subprocess.run(["bash", "-n", path], check=False).returncode, 0)
+
+    def test_materialize_forces_safe_path_for_existing_launchers(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            launcher = Path(temp_dir) / "unit.sh"
+            launcher.write_text("#!/usr/bin/env bash\nset -euo pipefail\npython -m dolma\n")
+            payload = _safe_path_launcher_payload(launcher, "123-test-run").decode()
+        self.assertIn("export PYTHONSAFEPATH=1", payload)
+        self.assertIn("cd /tmp", payload)
+        self.assertIn(
+            'export DOLMA_STATUS_ROOT="$HOME/dolma3p5-resharding-status/123-test-run"',
+            payload,
+        )
+
+    def test_materialize_defaults_to_a_complete_oe_other_worker_lifecycle(self):
+        args = build_materialize_parser().parse_args(["--all"])
+        self.assertEqual(args.project, "oe-other")
+        self.assertEqual(args.region, "us-east-1")
+        self.assertEqual(args.parallelism, 128)
+        self.assertEqual(args.provision_batch_size, 5)
+        self.assertEqual(args.provision_batch_delay_seconds, 3.0)
+        self.assertEqual(args.bootstrap_parallelism, 32)
+        self.assertEqual(args.readiness_poll_seconds, 10)
+        self.assertIsNone(args.instance_type)
+        self.assertEqual(args.storage_layout, "auto")
+        self.assertEqual(args.completion_poll_seconds, 30)
+        self.assertFalse(args.verbose)
+        self.assertFalse(args.preflight)
+        self.assertEqual(args.exclude_category, [])
+
+    def test_materialize_all_can_exclude_completed_categories(self):
+        rows = [
+            {
+                "unit_id": "00000001",
+                "leaf_id": "001:00",
+                "mix_name": "one:category",
+                "category_name": "default",
+            },
+            {
+                "unit_id": "00000002",
+                "leaf_id": "002:00",
+                "mix_name": "two:category",
+                "category_name": "high",
+            },
+            {
+                "unit_id": "00000003",
+                "leaf_id": "002:01",
+                "mix_name": "two:category",
+                "category_name": "mid",
+            },
+        ]
+        args = build_materialize_parser().parse_args(["--all", "--exclude-category", "one:category"])
+
+        label, selected = _select_units(args, rows)
+
+        self.assertEqual(label, "all-except-1-categories")
+        self.assertEqual([row["unit_id"] for row in selected], ["00000002", "00000003"])
+
+    def test_materialize_rejects_category_exclusion_without_all(self):
+        args = build_materialize_parser().parse_args(
+            [
+                "--category",
+                "one:category",
+                "--exclude-category",
+                "two:category",
+            ]
+        )
+
+        with self.assertRaisesRegex(PreparationError, "only be used with --all"):
+            _select_units(args, [])
+
+    @patch("scripts.dolma3p5_resharding.materialize.preflight_build")
+    def test_materialize_can_run_selection_aware_preflight(self, preflight):
+        args = SimpleNamespace(
+            profile="read-only",
+            region="us-east-1",
+            category="dolma3_finemath_v3:finemath::default",
+            unit=None,
+        )
+
+        selected = [{"unit_id": "00000136"}, {"unit_id": "00000137"}]
+        _run_selected_preflight(args, Path("/tmp/build"), selected)
+
+        inline_args = preflight.call_args.args[0]
+        self.assertEqual(inline_args.build, Path("/tmp/build"))
+        self.assertEqual(inline_args.profile, "read-only")
+        self.assertEqual(inline_args.region, "us-east-1")
+        self.assertEqual(
+            inline_args.category,
+            "dolma3_finemath_v3:finemath::default",
+        )
+        self.assertIsNone(inline_args.unit)
+        self.assertEqual(inline_args.selected_unit_ids, ("00000136", "00000137"))
+        self.assertIsNone(inline_args.max_workers)
+        self.assertTrue(inline_args.quiet)
+
+        parsed = build_materialize_parser().parse_args(["--all", "--execute", "--preflight"])
+        self.assertTrue(parsed.preflight)
+
+    def test_materialize_groups_units_by_planned_worker(self):
+        args = build_materialize_parser().parse_args(["--all"])
+        rows = [
+            {
+                "unit_id": "00000000",
+                "worker_instance_type": "i4i.2xlarge",
+                "worker_storage_layout": "single",
+                "worker_vcpus": "8",
+            },
+            {
+                "unit_id": "00000001",
+                "worker_instance_type": "i4i.8xlarge",
+                "worker_storage_layout": "raid0",
+                "worker_vcpus": "32",
+            },
+        ]
+
+        groups = _planned_worker_groups(args, rows)
+
+        self.assertEqual(
+            [(group.instance_type, group.storage_layout) for group in groups],
+            [("i4i.2xlarge", "single"), ("i4i.8xlarge", "raid0")],
+        )
+
+    def test_materialize_allocates_global_parallelism_across_worker_groups(self):
+        groups = [
+            PlannedWorkerGroup(
+                "i4i.2xlarge",
+                "single",
+                tuple({"unit_id": f"small-{index}"} for index in range(3)),
+            ),
+            PlannedWorkerGroup(
+                "i4i.8xlarge",
+                "raid0",
+                tuple({"unit_id": f"large-{index}"} for index in range(5)),
+            ),
+        ]
+
+        self.assertEqual(_worker_counts_for_groups(groups, 2), [1, 1])
+        self.assertEqual(_worker_counts_for_groups(groups, 4), [2, 2])
+        self.assertEqual(_worker_counts_for_groups(groups, 8), [3, 5])
+        with self.assertRaisesRegex(PreparationError, "must be at least 2"):
+            _worker_counts_for_groups(groups, 1)
+
+        large_group = [
+            PlannedWorkerGroup(
+                "i4i.2xlarge",
+                "single",
+                tuple({"unit_id": f"small-{index}"} for index in range(128)),
+            )
+        ]
+        self.assertEqual(_worker_counts_for_groups(large_group, 128), [90])
+
+    def test_materialize_batches_provider_create_requests(self):
+        self.assertEqual(_provision_batches(0, 5), ())
+        self.assertEqual(_provision_batches(3, 5), (3,))
+        self.assertEqual(_provision_batches(12, 5), (5, 5, 2))
+
+    @patch("scripts.dolma3p5_resharding.materialize.time.sleep")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    def test_materialize_batches_provider_resume_requests(self, run, sleep):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            profile=None,
+            parallelism=128,
+            provision_batch_size=5,
+            provision_batch_delay_seconds=3,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            ssh_key_path=None,
+        )
+
+        _resume_workers_in_batches(args, [f"i-{index:02d}" for index in range(12)])
+
+        self.assertEqual(
+            [lifecycle_call.args[0] for lifecycle_call in run.call_args_list],
+            [
+                "resume dolma3p5-14t-i4i-2xlarge-single batch 1/3",
+                "resume dolma3p5-14t-i4i-2xlarge-single batch 2/3",
+                "resume dolma3p5-14t-i4i-2xlarge-single batch 3/3",
+            ],
+        )
+        commands = [lifecycle_call.args[1] for lifecycle_call in run.call_args_list]
+        self.assertEqual([command.count("--instance-id") for command in commands], [5, 5, 2])
+        self.assertEqual(
+            [command[command.index("--parallelism") + 1] for command in commands],
+            ["5", "5", "2"],
+        )
+        self.assertEqual(sleep.call_args_list, [call(3.0), call(3.0)])
+
+    @patch("scripts.dolma3p5_resharding.materialize.time.sleep")
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    def test_materialize_retries_only_workers_that_remain_stopped(self, run, describe, sleep):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            profile=None,
+            parallelism=128,
+            provision_batch_size=5,
+            provision_batch_delay_seconds=3,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            ssh_key_path=None,
+        )
+        run.side_effect = [PreparationError("throttled"), None]
+        describe.return_value = [
+            ClusterInstance("i-running", "pending", "i4i.2xlarge", "oe-other"),
+            ClusterInstance("i-stopped", "stopped", "i4i.2xlarge", "oe-other"),
+        ]
+
+        _resume_workers_in_batches(args, ["i-running", "i-stopped"])
+
+        self.assertEqual(run.call_count, 2)
+        retry_command = run.call_args_list[1].args[1]
+        retry_ids = [
+            retry_command[index + 1] for index, value in enumerate(retry_command) if value == "--instance-id"
+        ]
+        self.assertEqual(retry_ids, ["i-stopped"])
+        sleep.assert_called_once_with(3.0)
+
+    @patch("scripts.dolma3p5_resharding.materialize.time.sleep")
+    @patch("scripts.dolma3p5_resharding.materialize._resume_worker_batch")
+    def test_materialize_resume_batch_size_is_global_across_worker_classes(self, resume_batch, sleep):
+        def group_args(instance_type, storage_layout):
+            return SimpleNamespace(
+                cluster="dolma3p5-14t",
+                project="oe-other",
+                region="us-east-1",
+                profile=None,
+                provision_batch_size=5,
+                provision_batch_delay_seconds=3,
+                instance_type=instance_type,
+                storage_layout=storage_layout,
+            )
+
+        small = group_args("i4i.2xlarge", "single")
+        medium = group_args("i4i.4xlarge", "single")
+        large = group_args("i4i.8xlarge", "raid0")
+        _resume_worker_groups_in_batches(
+            [
+                (small, ["i-s1", "i-s2", "i-s3"]),
+                (medium, ["i-m1", "i-m2", "i-m3"]),
+                (large, ["i-l1", "i-l2"]),
+            ],
+            console=Console(file=io.StringIO(), force_terminal=False),
+        )
+
+        calls_by_wave = {1: [], 2: []}
+        for resume_call in resume_batch.call_args_list:
+            stage = resume_call.kwargs["stage"]
+            wave = 1 if "batch 1/2" in stage else 2
+            calls_by_wave[wave].extend(resume_call.args[1])
+        self.assertEqual(len(calls_by_wave[1]), 5)
+        self.assertEqual(len(calls_by_wave[2]), 3)
+        self.assertEqual(
+            sorted(calls_by_wave[1] + calls_by_wave[2]),
+            [
+                "i-l1",
+                "i-l2",
+                "i-m1",
+                "i-m2",
+                "i-m3",
+                "i-s1",
+                "i-s2",
+                "i-s3",
+            ],
+        )
+        sleep.assert_called_once_with(3.0)
+
+    @patch("scripts.dolma3p5_resharding.materialize.time.sleep")
+    @patch("scripts.dolma3p5_resharding.materialize._run_compact_process")
+    def test_materialize_batches_failure_cleanup_requests(self, run, sleep):
+        run.return_value = 0
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            profile=None,
+            parallelism=128,
+            provision_batch_size=5,
+            provision_batch_delay_seconds=3,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            ssh_key_path=None,
+        )
+
+        _pause_workers_after_failure(args, [f"i-{index:02d}" for index in range(12)])
+
+        self.assertEqual(run.call_count, 3)
+        commands = [cleanup_call.args[1] for cleanup_call in run.call_args_list]
+        self.assertEqual([command.count("--instance-id") for command in commands], [5, 5, 2])
+        sleep.assert_has_calls([call(3.0), call(3.0)])
+
+    def test_materialize_queues_individual_units_largest_first(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            script_dir = Path(temp_dir)
+            rows = tuple(
+                {
+                    "unit_id": unit_id,
+                    "launcher_path": f"launchers/{unit_id}.sh",
+                    "planned_uint32_values": str(work),
+                    "estimated_work_uint32_values": str(work),
+                }
+                for unit_id, work in (("small", 10), ("large", 30), ("medium", 20))
+            )
+            for row in rows:
+                (script_dir / f"{row['unit_id']}.sh").write_text("#!/bin/bash\n")
+            group = MaterializationGroup(
+                args=SimpleNamespace(),
+                rows=rows,
+                script_dir=script_dir,
+                worker_count=2,
+            )
+
+            assignments = _stage_worker_assignments(group)
+
+            for assignment in assignments:
+                self.assertTrue(assignment.script_dir.is_dir())
+                scripts = list(assignment.script_dir.iterdir())
+                self.assertEqual(len(scripts), 1)
+                self.assertEqual(
+                    scripts[0].name,
+                    f"{assignment.rows[0]['unit_id']}.sh",
+                )
+
+        self.assertEqual(
+            [assignment.rows[0]["unit_id"] for assignment in assignments],
+            ["large", "medium", "small"],
+        )
+        self.assertTrue(all(len(assignment.rows) == 1 for assignment in assignments))
+
+    def test_dynamic_dispatch_does_not_spindown_between_units(self):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=1,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            ssh_key_path=None,
+        )
+        command = _map_command(
+            args,
+            Path("/tmp/unit.sh"),
+            ["i-first"],
+            spindown=False,
+        )
+
+        self.assertNotIn("--spindown", command)
+
+    @patch("scripts.dolma3p5_resharding.materialize._retag_cluster_instances")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    def test_materialize_creates_and_waits_for_the_exact_worker_count(self, describe, run, retag):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=8,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            root_storage_type="gp3",
+            root_storage_size=200,
+            ssh_key_path=None,
+            profile=None,
+        )
+        describe.side_effect = [
+            [],
+            [
+                ClusterInstance("i-second", "running", "i4i.2xlarge", "oe-other"),
+                ClusterInstance("i-first", "running", "i4i.2xlarge", "oe-other"),
+            ],
+        ]
+
+        self.assertEqual(_prepare_workers(args, 2), ["i-first", "i-second"])
+        self.assertEqual(
+            [call.args[0] for call in run.call_args_list],
+            ["create worker batch 1/1", "wait for workers"],
+        )
+        create_command = run.call_args_list[0].args[1]
+        self.assertEqual(create_command[create_command.index("--name") + 1], "dolma3p5-14t")
+        self.assertIn("--number", create_command)
+        self.assertEqual(create_command[create_command.index("--number") + 1], "2")
+        self.assertIn("--detach", create_command)
+        wait_command = run.call_args_list[1].args[1]
+        self.assertEqual(
+            wait_command[wait_command.index("--name") + 1],
+            "dolma3p5-14t-i4i-2xlarge-single",
+        )
+        self.assertEqual(wait_command.count("--instance-id"), 2)
+        retag.assert_called_once_with(
+            args,
+            ["i-first", "i-second"],
+            names={
+                "i-first": "dolma3p5-14t-0000",
+                "i-second": "dolma3p5-14t-0001",
+            },
+        )
+
+    @patch("scripts.dolma3p5_resharding.materialize._retag_cluster_instances")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    def test_materialize_launches_large_fleets_in_detached_batches(self, describe, run, retag):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=5,
+            provision_batch_size=2,
+            provision_batch_delay_seconds=0,
+            instance_type="i4i.2xlarge",
+            root_storage_type="gp3",
+            root_storage_size=200,
+            ssh_key_path=None,
+            profile=None,
+        )
+        workers = [ClusterInstance(f"i-{index}", "pending", "i4i.2xlarge", "oe-other") for index in range(5)]
+        describe.side_effect = [[], workers[:2], workers[:4], workers]
+
+        self.assertEqual(
+            _prepare_workers(args, 5),
+            [f"i-{index}" for index in range(5)],
+        )
+
+        self.assertEqual(
+            [call.args[0] for call in run.call_args_list],
+            [
+                "create worker batch 1/3",
+                "create worker batch 2/3",
+                "create worker batch 3/3",
+                "wait for workers",
+            ],
+        )
+        create_commands = [call.args[1] for call in run.call_args_list[:3]]
+        self.assertEqual(
+            [command[command.index("--number") + 1] for command in create_commands],
+            ["2", "2", "1"],
+        )
+        self.assertTrue(all("--detach" in command for command in create_commands))
+        self.assertEqual(
+            [command[command.index("--parallelism") + 1] for command in create_commands],
+            ["2", "2", "1"],
+        )
+        self.assertEqual(retag.call_count, 3)
+
+    @patch("scripts.dolma3p5_resharding.materialize.boto3.Session")
+    def test_materialize_retags_project_and_cluster_before_dispatch(self, session):
+        client = session.return_value.client.return_value
+        client.describe_instances.return_value = {
+            "Reservations": [
+                {
+                    "Instances": [
+                        {
+                            "InstanceId": "i-first",
+                            "Tags": [
+                                {
+                                    "Key": "project",
+                                    "Value": "dolma3p5-14t-i4i-2xlarge-single",
+                                },
+                                {"Key": "ai2-project", "Value": "oe-other"},
+                                {"Key": "cluster", "Value": "dolma3p5-14t"},
+                                {"Key": "Name", "Value": "dolma3p5-14t-0000"},
+                            ],
+                        }
+                    ]
+                }
+            ]
+        }
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            profile=None,
+        )
+
+        _retag_cluster_instances(
+            args,
+            ["i-first"],
+            names={"i-first": "dolma3p5-14t-0000"},
+        )
+
+        self.assertEqual(
+            client.create_tags.call_args_list[0],
+            call(
+                Resources=["i-first"],
+                Tags=[
+                    {
+                        "Key": "project",
+                        "Value": "dolma3p5-14t-i4i-2xlarge-single",
+                    },
+                    {"Key": "ai2-project", "Value": "oe-other"},
+                    {"Key": "cluster", "Value": "dolma3p5-14t"},
+                ],
+            ),
+        )
+
+    @patch("scripts.dolma3p5_resharding.materialize._retag_cluster_instances")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    def test_materialize_retags_reused_worker_before_resume(self, describe, run, retag):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=8,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            root_storage_type="gp3",
+            root_storage_size=200,
+            ssh_key_path=None,
+            profile=None,
+        )
+        describe.return_value = [
+            ClusterInstance(
+                "i-reused",
+                "stopped",
+                "i4i.2xlarge",
+                "oe-other",
+                "dolma3p5-14t-0003",
+            )
+        ]
+
+        self.assertEqual(_prepare_workers(args, 1), ["i-reused"])
+
+        retag.assert_called_once_with(args, ["i-reused"])
+        self.assertEqual(
+            [call.args[0] for call in run.call_args_list],
+            [
+                "resume dolma3p5-14t-i4i-2xlarge-single batch 1/1",
+                "wait for workers",
+            ],
+        )
+        for lifecycle_call in run.call_args_list:
+            command = lifecycle_call.args[1]
+            self.assertEqual(
+                command[command.index("--name") + 1],
+                "dolma3p5-14t-i4i-2xlarge-single",
+            )
+            self.assertEqual(command[command.index("--instance-id") + 1], "i-reused")
+
+    def test_existing_worker_commands_use_project_selector_and_explicit_ids(self):
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=8,
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+            ssh_key_path=None,
+        )
+        wait_command = _wait_command(args, ["i-first"])
+        map_command = _map_command(args, Path("/tmp/launchers"), ["i-first"])
+        log_command = _worker_log_command(args, ["i-first"], "test-run")
+        for command in (wait_command, map_command, log_command):
+            self.assertEqual(
+                command[command.index("--name") + 1],
+                "dolma3p5-14t-i4i-2xlarge-single",
+            )
+            self.assertEqual(command[command.index("--project") + 1], "oe-other")
+            self.assertEqual(command[command.index("--instance-id") + 1], "i-first")
+
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    def test_materialize_refuses_to_share_a_cluster_with_active_work(self, describe):
+        describe.return_value = [ClusterInstance("i-busy", "running", "i4i.2xlarge", "oe-other")]
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=8,
+            instance_type="i4i.2xlarge",
+            root_storage_type="gp3",
+            root_storage_size=200,
+            ssh_key_path=None,
+            profile=None,
+        )
+        with self.assertRaisesRegex(PreparationError, "already has active"):
+            _prepare_workers(args, 1)
+
+    @patch("scripts.dolma3p5_resharding.materialize._retag_cluster_instances")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    def test_materialize_allows_workers_owned_by_the_same_launch(self, describe, run, retag):
+        describe.return_value = [
+            ClusterInstance("i-owned", "running", "i4i.2xlarge", "oe-other"),
+            ClusterInstance("i-large", "stopped", "i4i.8xlarge", "oe-other"),
+        ]
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=8,
+            instance_type="i4i.8xlarge",
+            root_storage_type="gp3",
+            root_storage_size=200,
+            ssh_key_path=None,
+            profile=None,
+        )
+
+        self.assertEqual(
+            _prepare_workers(args, 1, owned_instance_ids=["i-owned"]),
+            ["i-large"],
+        )
+        retag.assert_called_once_with(args, ["i-large"])
+        self.assertEqual(
+            [call.args[0] for call in run.call_args_list],
+            ["resume dolma3p5-14t batch 1/1", "wait for workers"],
+        )
+
+    @patch("scripts.dolma3p5_resharding.materialize._retag_cluster_instances")
+    @patch("scripts.dolma3p5_resharding.materialize._run_lifecycle_command")
+    @patch("scripts.dolma3p5_resharding.materialize._describe_cluster_instances")
+    def test_materialize_can_defer_reused_worker_resume(self, describe, run, retag):
+        describe.return_value = [ClusterInstance("i-reused", "stopped", "i4i.2xlarge", "oe-other")]
+        args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            region="us-east-1",
+            parallelism=2,
+            instance_type="i4i.2xlarge",
+            root_storage_type="gp3",
+            root_storage_size=200,
+            ssh_key_path=None,
+            profile=None,
+        )
+        resume_ids = []
+
+        self.assertEqual(
+            _prepare_workers(
+                args,
+                1,
+                wait_for_ready=False,
+                deferred_resume_ids=resume_ids,
+            ),
+            ["i-reused"],
+        )
+
+        self.assertEqual(resume_ids, ["i-reused"])
+        retag.assert_called_once_with(args, ["i-reused"])
+        run.assert_not_called()
+
+    @patch("scripts.dolma3p5_resharding.materialize._pause_workers_after_failure")
+    @patch("scripts.dolma3p5_resharding.materialize._verify_materialized_units")
+    @patch("scripts.dolma3p5_resharding.materialize._wait_for_workers_to_stop")
+    @patch("scripts.dolma3p5_resharding.materialize._stop_worker_after_work")
+    @patch("scripts.dolma3p5_resharding.materialize._dispatch_assignment_to_worker")
+    @patch("scripts.dolma3p5_resharding.materialize._bootstrap_and_dispatch_worker")
+    @patch("scripts.dolma3p5_resharding.materialize._ready_worker_ids")
+    @patch("scripts.dolma3p5_resharding.materialize._stage_worker_assignments")
+    @patch("scripts.dolma3p5_resharding.materialize._resume_worker_groups_in_batches")
+    @patch("scripts.dolma3p5_resharding.materialize._prepare_workers")
+    def test_materialize_dispatches_workers_as_each_becomes_ready(
+        self,
+        prepare,
+        resume_workers,
+        stage_assignments,
+        ready_worker_ids,
+        bootstrap,
+        dispatch,
+        stop_worker,
+        monitor,
+        verify,
+        pause,
+    ):
+        events = []
+
+        def prepare_group(
+            group_args,
+            worker_count,
+            *,
+            owned_instance_ids=(),
+            wait_for_ready=True,
+            delay_after_last_batch=False,
+            deferred_resume_ids=None,
+        ):
+            instance_id = "i-small" if group_args.instance_type == "i4i.2xlarge" else "i-large"
+            events.append(
+                (
+                    "prepare",
+                    group_args.instance_type,
+                    tuple(owned_instance_ids),
+                    wait_for_ready,
+                    delay_after_last_batch,
+                )
+            )
+            if deferred_resume_ids is not None:
+                deferred_resume_ids.append(instance_id)
+            return [instance_id]
+
+        prepare.side_effect = prepare_group
+        selected = [
+            {
+                "unit_id": "small",
+                "planned_uint32_values": "10",
+                "estimated_work_uint32_values": "20",
+            },
+            {
+                "unit_id": "large",
+                "planned_uint32_values": "20",
+                "estimated_work_uint32_values": "40",
+            },
+        ]
+        small_args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+        )
+        large_args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            instance_type="i4i.8xlarge",
+            storage_layout="raid0",
+        )
+        groups = [
+            MaterializationGroup(
+                args=small_args,
+                rows=(selected[0],),
+                script_dir=Path("/tmp/small"),
+                worker_count=1,
+            ),
+            MaterializationGroup(
+                args=large_args,
+                rows=(selected[1],),
+                script_dir=Path("/tmp/large"),
+                worker_count=1,
+            ),
+        ]
+        stage_assignments.side_effect = [
+            (WorkerAssignment(groups[0], groups[0].rows, Path("/tmp/small-assignment")),),
+            (WorkerAssignment(groups[1], groups[1].rows, Path("/tmp/large-assignment")),),
+        ]
+        ready_worker_ids.side_effect = [{"i-small"}, {"i-large"}]
+
+        def dispatch_worker(dispatch_args, instance_id, assignment, stages, stage_lock, console):
+            events.append(("dispatch", instance_id, assignment.rows[0]["unit_id"]))
+            with stage_lock:
+                stages[instance_id] = "materializing"
+
+        bootstrap.side_effect = dispatch_worker
+
+        def monitor_workers(*monitor_args, **monitor_kwargs):
+            deadline = time.monotonic() + 2
+            while True:
+                with monitor_kwargs["stage_lock"]:
+                    stages = dict(monitor_kwargs["worker_stages"])
+                if set(stages.values()) == {"materializing"}:
+                    break
+                if time.monotonic() >= deadline:
+                    self.fail("workers were not dispatched")
+                time.sleep(0.01)
+            with monitor_kwargs["unit_status_lock"]:
+                monitor_kwargs["unit_statuses"].update(
+                    {
+                        ("i-small", "small"): "succeeded",
+                        ("i-large", "large"): "succeeded",
+                    }
+                )
+            monitor_kwargs["status_changed"].set()
+            self.assertTrue(monitor_kwargs["all_dispatched"].wait(timeout=2))
+            events.append(("monitor-complete",))
+
+        monitor.side_effect = monitor_workers
+        verify.side_effect = lambda *verify_args, **verify_kwargs: events.append(("verify",))
+        args = SimpleNamespace(
+            verbose=True,
+            bootstrap_parallelism=2,
+            readiness_poll_seconds=1,
+        )
+
+        _execute_materialization_groups(args, groups, selected, "test-run")
+
+        self.assertEqual(
+            events[:2],
+            [
+                ("prepare", "i4i.2xlarge", (), False, True),
+                ("prepare", "i4i.8xlarge", ("i-small",), False, False),
+            ],
+        )
+        resume_workers.assert_called_once_with(
+            [(small_args, ["i-small"]), (large_args, ["i-large"])],
+            console=ANY,
+        )
+        self.assertEqual(
+            [event for event in events if event[0] == "dispatch"],
+            [("dispatch", "i-small", "small"), ("dispatch", "i-large", "large")],
+        )
+        self.assertEqual(
+            ready_worker_ids.call_args_list,
+            [call(args, ["i-large", "i-small"]), call(args, ["i-large"])],
+        )
+        monitor.assert_called_once()
+        self.assertEqual(stop_worker.call_count, 2)
+        dispatch.assert_not_called()
+        verify.assert_called_once_with(args, selected)
+        pause.assert_not_called()
+
+    @patch("scripts.dolma3p5_resharding.materialize._pause_workers_after_failure")
+    @patch("scripts.dolma3p5_resharding.materialize._verify_materialized_units")
+    @patch("scripts.dolma3p5_resharding.materialize._wait_for_workers_to_stop")
+    @patch("scripts.dolma3p5_resharding.materialize._stop_worker_after_work")
+    @patch("scripts.dolma3p5_resharding.materialize._dispatch_assignment_to_worker")
+    @patch("scripts.dolma3p5_resharding.materialize._bootstrap_and_dispatch_worker")
+    @patch("scripts.dolma3p5_resharding.materialize._ready_worker_ids")
+    @patch("scripts.dolma3p5_resharding.materialize._stage_worker_assignments")
+    @patch("scripts.dolma3p5_resharding.materialize._resume_worker_groups_in_batches")
+    @patch("scripts.dolma3p5_resharding.materialize._prepare_workers")
+    def test_materialize_reuses_free_workers_for_the_next_unit(
+        self,
+        prepare,
+        resume_workers,
+        stage_assignments,
+        ready_worker_ids,
+        bootstrap,
+        dispatch,
+        stop_worker,
+        monitor,
+        verify,
+        pause,
+    ):
+        rows = tuple(
+            {
+                "unit_id": f"unit-{index}",
+                "planned_uint32_values": "10",
+                "estimated_work_uint32_values": "10",
+            }
+            for index in range(4)
+        )
+        group_args = SimpleNamespace(
+            cluster="dolma3p5-14t",
+            project="oe-other",
+            instance_type="i4i.2xlarge",
+            storage_layout="single",
+        )
+        group = MaterializationGroup(
+            args=group_args,
+            rows=rows,
+            script_dir=Path("/tmp/dynamic"),
+            worker_count=2,
+        )
+        assignments = tuple(WorkerAssignment(group, (row,), Path(f"/tmp/{row['unit_id']}.sh")) for row in rows)
+        stage_assignments.return_value = assignments
+
+        def prepare_workers(*_args, deferred_resume_ids=None, **_kwargs):
+            if deferred_resume_ids is not None:
+                deferred_resume_ids.extend(["i-first", "i-second"])
+            return ["i-first", "i-second"]
+
+        prepare.side_effect = prepare_workers
+        ready_worker_ids.return_value = {"i-first", "i-second"}
+        assigned: list[tuple[str, str]] = []
+        assigned_lock = Lock()
+
+        def record_assignment(instance_id, assignment, stages, stage_lock):
+            with assigned_lock:
+                assigned.append((instance_id, assignment.rows[0]["unit_id"]))
+            with stage_lock:
+                stages[instance_id] = "materializing"
+
+        def bootstrap_worker(_args, instance_id, assignment, stages, stage_lock, _console):
+            record_assignment(instance_id, assignment, stages, stage_lock)
+
+        def dispatch_worker(instance_id, assignment, stages, stage_lock, _console):
+            record_assignment(instance_id, assignment, stages, stage_lock)
+
+        bootstrap.side_effect = bootstrap_worker
+        dispatch.side_effect = dispatch_worker
+
+        def wait_for_assignment_count(count):
+            deadline = time.monotonic() + 2
+            while True:
+                with assigned_lock:
+                    if len(assigned) >= count:
+                        return list(assigned)
+                if time.monotonic() >= deadline:
+                    self.fail(f"only {len(assigned)} assignments were dispatched")
+                time.sleep(0.01)
+
+        def monitor_workers(*_args, **monitor_kwargs):
+            initial = wait_for_assignment_count(2)
+            with monitor_kwargs["unit_status_lock"]:
+                monitor_kwargs["unit_statuses"].update(
+                    {(instance_id, unit_id): "succeeded" for instance_id, unit_id in initial}
+                )
+            monitor_kwargs["status_changed"].set()
+
+            all_assignments = wait_for_assignment_count(4)
+            with monitor_kwargs["unit_status_lock"]:
+                monitor_kwargs["unit_statuses"].update(
+                    {(instance_id, unit_id): "succeeded" for instance_id, unit_id in all_assignments[2:]}
+                )
+            monitor_kwargs["status_changed"].set()
+            self.assertTrue(monitor_kwargs["all_dispatched"].wait(timeout=2))
+
+        monitor.side_effect = monitor_workers
+        args = SimpleNamespace(
+            verbose=False,
+            bootstrap_parallelism=2,
+            readiness_poll_seconds=1,
+        )
+
+        _execute_materialization_groups(args, [group], rows, "test-run")
+
+        with assigned_lock:
+            assigned_units = [unit_id for _, unit_id in assigned]
+        self.assertEqual(len(assigned_units), 4)
+        self.assertEqual(set(assigned_units), {row["unit_id"] for row in rows})
+        self.assertEqual(bootstrap.call_count, 2)
+        self.assertEqual(dispatch.call_count, 2)
+        self.assertEqual(stop_worker.call_count, 2)
+        resume_workers.assert_called_once()
+        verify.assert_called_once_with(args, rows)
+        pause.assert_not_called()
+
+    @patch("scripts.dolma3p5_resharding.materialize._pause_workers_after_failure")
+    @patch("scripts.dolma3p5_resharding.materialize._stage_worker_assignments")
+    @patch("scripts.dolma3p5_resharding.materialize._prepare_workers")
+    def test_materialize_cleans_up_prepared_groups_if_later_provisioning_fails(
+        self, prepare, stage_assignments, pause
+    ):
+        prepare.side_effect = [["i-small"], PreparationError("create failed")]
+        args = SimpleNamespace(verbose=False)
+        selected = [{"unit_id": "small"}, {"unit_id": "large"}]
+        small_args = SimpleNamespace(instance_type="i4i.2xlarge")
+        groups = [
+            MaterializationGroup(
+                args=small_args,
+                rows=(selected[0],),
+                script_dir=Path("/tmp/small"),
+                worker_count=1,
+            ),
+            MaterializationGroup(
+                args=SimpleNamespace(instance_type="i4i.8xlarge"),
+                rows=(selected[1],),
+                script_dir=Path("/tmp/large"),
+                worker_count=1,
+            ),
+        ]
+        stage_assignments.side_effect = [
+            (WorkerAssignment(groups[0], groups[0].rows, Path("/tmp/small-worker")),),
+            (WorkerAssignment(groups[1], groups[1].rows, Path("/tmp/large-worker")),),
+        ]
+
+        with self.assertRaisesRegex(PreparationError, "create failed"):
+            _execute_materialization_groups(args, groups, selected, "test-run")
+
+        pause.assert_called_once_with(small_args, ["i-small"])
+
+    def test_region_defaults_to_us_east_1_and_allows_override(self):
+        self.assertEqual(normalize_region(None), DEFAULT_REGION)
+        self.assertEqual(normalize_region(""), DEFAULT_REGION)
+        self.assertEqual(normalize_region("  "), DEFAULT_REGION)
+        self.assertEqual(normalize_region(" us-west-2 "), "us-west-2")
+
+    def test_source_layout_preserves_exact_source_directory(self):
+        self.assertEqual(
+            _source_relative_directory(
+                "preprocessed/dolma3-0625/v0.1-official/allenai/" "dolma3-tokenizer/finemath-3plus/000000.npy"
+            ),
+            "dolma3-0625/v0.1-official/allenai/dolma3-tokenizer/finemath-3plus",
+        )
+        self.assertEqual(
+            _source_relative_directory(
+                "preprocessed/cc_all_dressed/all_dressed_v5/topic/health/"
+                "vigintile_0016/allenai/dolma2-tokenizer/000000.npy"
+            ),
+            "cc_all_dressed/all_dressed_v5/topic/health/" "vigintile_0016/allenai/dolma2-tokenizer",
+        )
+
+    def test_category_output_directory_replaces_varying_lower_group(self):
+        objects = [
+            {
+                "key": (
+                    "preprocessed/cc_all_dressed/all_dressed_v5/topic/health/"
+                    f"vigintile_{index:04d}/allenai/dolma2-tokenizer/000000.npy"
+                )
+            }
+            for index in range(16, 20)
+        ]
+        self.assertEqual(
+            _category_output_directory(objects, "high"),
+            "cc_all_dressed/all_dressed_v5/topic/health/high/allenai/dolma2-tokenizer",
+        )
+
+    def setUp(self):
+        self.temp_dir = tempfile.TemporaryDirectory()
+        self.root = Path(self.temp_dir.name)
+        self.mix = self.root / "mix.yaml"
+        self.catalog = self.root / "catalog.csv"
+        mix = {
+            "mix": [
+                {
+                    "name": "catalog-source:topic",
+                    "weight": 0.5,
+                    "categories": [
+                        {
+                            "name": "default",
+                            "weight": 1.0,
+                            "paths": ["dolma3p5_pool/catalog-source/topic/allenai/tokenizer/*.npy"],
+                            "repetition_factor": -1.0,
+                        },
+                        {
+                            "name": "dropped",
+                            "weight": 0.0,
+                            "paths": ["dolma3p5_pool/catalog-source/topic/vigintile_0000/allenai/tokenizer/*.npy"],
+                            "repetition_factor": -1.0,
+                        },
+                    ],
+                },
+                {
+                    "name": "the-stack-v2:Tcl",
+                    "weight": 0.25,
+                    "categories": [
+                        {
+                            "name": "high",
+                            "weight": 1.0,
+                            "paths": [
+                                "dolma3p5_pool/the-stack-v2/data/Tcl/quality_p95/allenai/tokenizer/*.npy",
+                                "dolma3p5_pool/the-stack-v2/data/Tcl/quality_p95/allenai/tokenizer/*.npy",
+                            ],
+                            "repetition_factor": -1.0,
+                        }
+                    ],
+                },
+                {
+                    "name": "direct-source:default",
+                    "weight": 0.25,
+                    "categories": [
+                        {
+                            "name": "default",
+                            "weight": 1.0,
+                            "paths": ["preprocessed/direct-source/allenai/tokenizer/*.npy"],
+                            "repetition_factor": -1.0,
+                        }
+                    ],
+                },
+            ]
+        }
+        self.mix.write_text(yaml.safe_dump(mix, sort_keys=False))
+        self.catalog.write_text(
+            "ai2-llm,preprocessed/catalog-source/topic/allenai/tokenizer/0000.npy\n"
+            "ai2-llm,preprocessed/catalog-source/topic/allenai/tokenizer/0000.csv.gz\n"
+            "ai2-llm,preprocessed/catalog-source/topic/vigintile_0000/allenai/tokenizer/0000.npy\n"
+            "ai2-llm,preprocessed/the-stack-v2/data/Tcl/quality_p95/allenai/tokenizer/0000.npy\n"
+        )
+        self.settings = self.root / "settings.yaml"
+        self.settings.write_text(
+            yaml.safe_dump(
+                {
+                    "worker_instance_grid": [
+                        {
+                            "instance_type": "i4i.32xlarge",
+                            "vcpus": 128,
+                            "local_nvme_devices": 8,
+                            "local_nvme_bytes": 30_000_000_000_000,
+                            "max_estimated_work_uint32_values": None,
+                        }
+                    ]
+                },
+                sort_keys=False,
+            )
+        )
+        self.build = self.root / "build"
+
+    def tearDown(self):
+        self.temp_dir.cleanup()
+
+    def _plan(self):
+        plan_build(
+            argparse.Namespace(
+                mix=self.mix,
+                catalog=self.catalog,
+                settings=self.settings,
+                output=self.build,
+            )
+        )
+
+    def _write_inventory(self):
+        listing = self.root / "listing.jsonl"
+        with listing.open("x") as f:
+            for uri, size in self._inventory_objects().items():
+                f.write(json.dumps(self._inventory_record(uri, size)) + "\n")
+        phase = self.build / "01-plan/inventory"
+        phase.mkdir()
+        objects, errors = _parse_s5cmd_jsonl(listing)
+        self.assertFalse(errors)
+        _finalize_inventory(
+            self.build,
+            phase,
+            objects,
+            client=MagicMock(),
+            max_workers=1,
+        )
+
+    @staticmethod
+    def _inventory_objects():
+        return {
+            "s3://ai2-llm/preprocessed/catalog-source/topic/allenai/tokenizer/0000.npy": 400,
+            "s3://ai2-llm/preprocessed/catalog-source/topic/allenai/tokenizer/0000.csv.gz": 20,
+            "s3://ai2-llm/preprocessed/catalog-source/topic/vigintile_0000/allenai/tokenizer/0000.npy": 200,
+            "s3://ai2-llm/preprocessed/catalog-source/topic/vigintile_0000/allenai/tokenizer/0000.csv.gz": 16,
+            "s3://ai2-llm/preprocessed/the-stack-v2/data/Tcl/quality_p95/allenai/tokenizer/0000.npy": 800,
+            "s3://ai2-llm/preprocessed/the-stack-v2/data/Tcl/quality_p95/allenai/tokenizer/0000.csv.gz": 24,
+            "s3://ai2-llm/preprocessed/direct-source/allenai/tokenizer/0000.npy": 1200,
+            "s3://ai2-llm/preprocessed/direct-source/allenai/tokenizer/0000.csv.gz": 28,
+        }
+
+    @staticmethod
+    def _inventory_record(uri, size):
+        return {
+            "key": uri,
+            "etag": "test-etag",
+            "last_modified": "2026-01-01T00:00:00Z",
+            "type": "file",
+            "size": size,
+            "storage_class": "STANDARD",
+        }
+
+    def test_inventory_uses_s5cmd(self):
+        self._plan()
+        captured_command = []
+
+        def run_collector(command, raw_output, environment, status):
+            captured_command.extend(command)
+            self.assertIn("AWS_PROFILE", environment)
+            with raw_output.open("x") as stdout:
+                for uri, size in self._inventory_objects().items():
+                    stdout.write(json.dumps(self._inventory_record(uri, size)) + "\n")
+            status("Running: 8 JSON records received, 1s elapsed")
+            return SimpleNamespace(
+                returncode=0,
+                stderr="",
+                output_records=len(self._inventory_objects()),
+                elapsed_seconds=1.0,
+            )
+
+        session = MagicMock()
+        output = io.StringIO()
+        with redirect_stdout(output):
+            with (
+                patch(
+                    "scripts.dolma3p5_resharding.workflow.shutil.which",
+                    return_value="/usr/bin/s5cmd",
+                ),
+                patch(
+                    "scripts.dolma3p5_resharding.workflow._run_s5cmd_inventory",
+                    side_effect=run_collector,
+                ),
+                patch(
+                    "scripts.dolma3p5_resharding.workflow.boto3.Session",
+                    return_value=session,
+                ),
+            ):
+                inventory_args = argparse.Namespace(
+                    build=self.build,
+                    profile="read-only",
+                    region="us-west-2",
+                    max_workers=2,
+                )
+                collect_inventory(inventory_args)
+                collect_inventory(inventory_args)
+
+        self.assertEqual(captured_command[0], "s5cmd")
+        self.assertEqual(captured_command.count("s5cmd"), 2)
+        collector = json.loads((self.build / "01-plan/inventory/collector.json").read_text())
+        self.assertEqual(collector["collector"], "s5cmd")
+        self.assertEqual(collector["output_records"], 8)
+        self.assertTrue((self.build / "01-plan/inventory/raw-listings.jsonl").is_file())
+        status_output = output.getvalue()
+        self.assertIn("[inventory 1/4] Bulk listing:", status_output)
+        self.assertIn("[inventory 1/4] Running:", status_output)
+        self.assertIn("[inventory 2/4] Resolving required objects for", status_output)
+        self.assertIn("[inventory 3/4] Validation:", status_output)
+        self.assertIn("[inventory 3/4] Estimated source tokens:", status_output)
+        self.assertIn("[inventory 4/4] PASS:", status_output)
+
+    def test_inventory_requires_s5cmd_without_replacing_existing_artifacts(self):
+        self._plan()
+        inventory_phase = self.build / "01-plan/inventory"
+        inventory_phase.mkdir()
+        existing = inventory_phase / "keep.txt"
+        existing.write_text("existing inventory")
+
+        with (
+            patch(
+                "scripts.dolma3p5_resharding.workflow.shutil.which",
+                return_value=None,
+            ),
+            self.assertRaisesRegex(PreparationError, "s5cmd is required"),
+        ):
+            collect_inventory(
+                argparse.Namespace(
+                    build=self.build,
+                    profile=None,
+                    region=None,
+                    max_workers=None,
+                )
+            )
+        self.assertEqual(existing.read_text(), "existing inventory")
+
+    def test_catalog_resolves_yaml_wildcard_against_live_listing(self):
+        self._plan()
+        objects = self._inventory_objects() | {
+            "s3://ai2-llm/preprocessed/catalog-source/topic/allenai/tokenizer/0001.npy": 100,
+            "s3://ai2-llm/preprocessed/catalog-source/topic/allenai/tokenizer/0001.csv.gz": 12,
+        }
+        listing = self.root / "expanded-listing.jsonl"
+        with listing.open("x") as handle:
+            for uri, size in objects.items():
+                handle.write(json.dumps(self._inventory_record(uri, size)) + "\n")
+        listed, errors = _parse_s5cmd_jsonl(listing)
+        self.assertFalse(errors)
+        phase = self.build / "01-plan/inventory"
+        phase.mkdir()
+        summary = _finalize_inventory(
+            self.build,
+            phase,
+            listed,
+            client=MagicMock(),
+            max_workers=1,
+        )
+        with (phase / "required-objects.csv").open() as handle:
+            required = list(csv.DictReader(handle))
+        catalog_topic = [
+            row for row in required if row["leaf_id"] == "000:00" and row["category_name"] == "default"
+        ]
+        self.assertEqual(
+            [row["key"].rsplit("/", 1)[-1] for row in catalog_topic],
+            ["0000.npy", "0001.npy"],
+        )
+        self.assertEqual(summary["source_uint32_values"], 675)
+
+    def test_sampling_rate_limit_blocks_proposal(self):
+        self._plan()
+        build_manifest_path = self.build / "build.json"
+        build_manifest = json.loads(build_manifest_path.read_text())
+        build_manifest["settings"]["maximum_expected_upsample_rate"] = 7.2
+        build_manifest_path.write_text(json.dumps(build_manifest))
+        listing = self.root / "limited-listing.jsonl"
+        with listing.open("x") as handle:
+            for uri, size in self._inventory_objects().items():
+                handle.write(json.dumps(self._inventory_record(uri, size)) + "\n")
+        listed, errors = _parse_s5cmd_jsonl(listing)
+        self.assertFalse(errors)
+        phase = self.build / "01-plan/inventory"
+        phase.mkdir()
+        with self.assertRaisesRegex(PreparationError, "Inventory validation failed"):
+            _finalize_inventory(
+                self.build,
+                phase,
+                listed,
+                client=MagicMock(),
+                max_workers=1,
+            )
+        summary = json.loads((phase / "inventory-summary.json").read_text())
+        self.assertEqual(summary["sampling_rate_limit_failures"], 3)
+        with (phase / "sampling-rate-audit.csv").open() as handle:
+            audit = list(csv.DictReader(handle))
+        self.assertEqual(
+            sum(row["status"] == "above_expected_maximum" for row in audit),
+            3,
+        )
+        inventory_report = (phase / "report.html").read_text()
+        self.assertIn("Source-size consistency check failed", inventory_report)
+        self.assertIn("Proposal generation is blocked", inventory_report)
+        with self.assertRaisesRegex(PreparationError, "blocking validation failures"):
+            propose_configs(
+                argparse.Namespace(
+                    build=self.build,
+                    destination_root="s3://test-bucket/new-datasets/dolma3p5",
+                    local_temp_root=str(self.root / "temp-base"),
+                    max_unit_working_bytes=20_000_000_000_000,
+                )
+            )
+
+    def test_plan_reports_resolution_counts_without_chart_artifacts(self):
+        output = io.StringIO()
+        with redirect_stdout(output):
+            self._plan()
+        self.assertIn(
+            "Plan summary: 3 catalog NPY matches, 1 direct S3 pattern, " "1 correction, 0 blocking failures",
+            output.getvalue(),
+        )
+        self.assertFalse((self.build / "01-plan/resolution/plots/resolution-counts.svg").exists())
+        self.assertFalse((self.build / "01-plan/resolution/plot-data/resolution-counts.csv").exists())
+
+    def test_plan_command_builds_resolution_inventory_and_execution_together(self):
+        from scripts.dolma3p5_resharding import plan as plan_command
+
+        output = self.root / "combined-plan"
+        with (
+            patch.object(plan_command.shutil, "which", return_value="/usr/bin/s5cmd"),
+            patch.object(plan_command, "plan_build") as build_paths,
+            patch.object(plan_command, "collect_inventory") as inventory,
+            patch.object(plan_command, "propose_configs") as execution,
+            patch.object(
+                sys,
+                "argv",
+                [
+                    "plan.py",
+                    "--mix",
+                    str(self.mix),
+                    "--catalog",
+                    str(self.catalog),
+                    "--output",
+                    str(output),
+                    "--profile",
+                    "read-only",
+                    "--destination-root",
+                    "s3://test-bucket/new-datasets/dolma3p5",
+                    "--local-temp-root",
+                    "/mnt/dolma/dolma3p5-resharding",
+                    "--max-unit-working-bytes",
+                    "1500000000000",
+                ],
+            ),
+        ):
+            plan_command.main()
+
+        build_paths.assert_called_once()
+        inventory.assert_called_once()
+        execution.assert_called_once()
+        inventory_args = inventory.call_args.args[0]
+        self.assertEqual(inventory_args.build, output)
+        self.assertEqual(inventory_args.profile, "read-only")
+        execution_args = execution.call_args.args[0]
+        self.assertEqual(execution_args.build, output)
+        self.assertEqual(
+            execution_args.destination_root,
+            "s3://test-bucket/new-datasets/dolma3p5",
+        )
+        self.assertEqual(execution_args.local_temp_root, Path("/mnt/dolma/dolma3p5-resharding"))
+        self.assertEqual(execution_args.max_unit_working_bytes, 1_500_000_000_000)
+
+    def test_catalog_paths_are_decoded_to_literal_s3_keys(self):
+        catalog = self.root / "encoded-catalog.csv"
+        catalog.write_text("ai2-llm,preprocessed/the-stack-v2/C%2B%2B/0000.npy\n")
+
+        rows = _load_catalog(catalog)
+
+        self.assertEqual(rows[0]["key"], "preprocessed/the-stack-v2/C++/0000.npy")
+
+    def test_execution_units_support_exact_category_and_unit_selection(self):
+        rows = [
+            {
+                "unit_id": "finemath-0001",
+                "leaf_id": "060:00",
+                "mix_name": "dolma3_finemath_v3:finemath",
+                "category_name": "default",
+            },
+            {
+                "unit_id": "finemath-0002",
+                "leaf_id": "060:00",
+                "mix_name": "dolma3_finemath_v3:finemath",
+                "category_name": "default",
+            },
+            {
+                "unit_id": "stack-0001",
+                "leaf_id": "061:00",
+                "mix_name": "the-stack-v2:Python",
+                "category_name": "high",
+            },
+        ]
+        selected_mix = _filter_execution_units(rows, category="dolma3_finemath_v3:finemath")
+        selected_leaf = _filter_execution_units(rows, category="060:00")
+        selected_full = _filter_execution_units(
+            rows,
+            category="the-stack-v2:Python::high",
+        )
+        selected_unit = _filter_execution_units(rows, unit="finemath-0002")
+
+        self.assertEqual([row["unit_id"] for row in selected_mix], ["finemath-0001", "finemath-0002"])
+        self.assertEqual(_unit_selection_digest(selected_mix), _unit_selection_digest(selected_leaf))
+        self.assertEqual([row["unit_id"] for row in selected_full], ["stack-0001"])
+        self.assertEqual([row["unit_id"] for row in selected_unit], ["finemath-0002"])
+
+    def test_s5cmd_listing_commands_use_literal_s3_keys(self):
+        mix = self.root / "encoded-mix.yaml"
+        catalog = self.root / "encoded-command-catalog.csv"
+        build = self.root / "encoded-command-build"
+        yaml_path = "dolma3p5_pool/the-stack-v2/C++/quality_p95/allenai/dolma2-tokenizer/*.npy"
+        mix.write_text(
+            yaml.safe_dump(
+                {
+                    "mix": [
+                        {
+                            "name": "the-stack-v2:C++",
+                            "weight": 1.0,
+                            "categories": [
+                                {
+                                    "name": "high",
+                                    "weight": 1.0,
+                                    "paths": [yaml_path],
+                                    "repetition_factor": -1.0,
+                                }
+                            ],
+                        }
+                    ]
+                },
+                sort_keys=False,
+            )
+        )
+        catalog.write_text(
+            "ai2-llm,preprocessed/the-stack-v2/C%2B%2B/quality_p95/" "allenai/dolma2-tokenizer/0000.npy\n"
+        )
+
+        plan_build(
+            argparse.Namespace(
+                mix=mix,
+                catalog=catalog,
+                settings=None,
+                output=build,
+            )
+        )
+
+        commands = (build / "01-plan/resolution/bulk-listing-commands.txt").read_text()
+        self.assertIn("/C++/quality_p95/", commands)
+        self.assertNotIn("%2B", commands)
+
+    def test_end_to_end_preparation_phases_are_replaceable(self):
+        mix_before = self.mix.read_bytes()
+        catalog_before = self.catalog.read_bytes()
+        self._plan()
+        plan_report = (self.build / "01-plan/resolution/report.html").read_text()
+        self.assertIn("Dolma 3.5 Target Allocation and Source Path Plan", plan_report)
+        self.assertIn("Materialized output target:", plan_report)
+        self.assertNotIn("S3 source volume:", plan_report)
+        self.assertIn('data-detail="plan-family-detail-', plan_report)
+        self.assertIn('class="subcategory-row', plan_report)
+        self.assertIn('class="subcategory-detail"', plan_report)
+        self.assertIn('class="category-grid"', plan_report)
+        self.assertIn("matched NPY", plan_report)
+        self.assertIn("topic", plan_report)
+        self.assertIn(
+            "s3://ai2-llm/preprocessed/catalog-source/topic/allenai/tokenizer/0000.npy",
+            plan_report,
+        )
+        self.assertIn(
+            "s3://ai2-llm/preprocessed/direct-source/allenai/tokenizer/*.npy",
+            plan_report,
+        )
+        self.assertNotIn(
+            "dolma3p5_pool/catalog-source/topic/allenai/tokenizer/*.npy",
+            plan_report,
+        )
+        with (self.build / "01-plan/resolution/normalized-paths.csv").open() as f:
+            self.assertNotIn("resolution_route", csv.DictReader(f).fieldnames)
+        with (self.build / "01-plan/resolution/listing-plan.csv").open() as f:
+            self.assertNotIn("resolution_routes", csv.DictReader(f).fieldnames)
+        plan_target_plot = (self.build / "01-plan/resolution/plots/target-mix.svg").read_text()
+        self.assertIn("Total target: 14T tokens (14,000,000,000,000)", plan_target_plot)
+        self.assertIn("50.00% · 7T tokens", plan_target_plot)
+        self.assertNotIn('text-anchor="end"', plan_target_plot)
+        with (self.build / "01-plan/resolution/corrections.csv").open() as f:
+            corrections = list(csv.DictReader(f))
+        self.assertEqual(len(corrections), 1)
+        self.assertIn("quality_p95", corrections[0]["path"])
+        with (self.build / "01-plan/resolution/direct-s3-patterns.csv").open() as f:
+            direct = list(csv.DictReader(f))
+        self.assertEqual(len(direct), 1)
+        finder_metadata = self.build / ".DS_Store"
+        finder_metadata.write_bytes(b"preserve benign metadata")
+        legacy_proposal = self.build / "03-proposal"
+        legacy_proposal.mkdir()
+        (legacy_proposal / "old-generated-artifact.txt").write_text("replaceable")
+        self._plan()
+        self.assertEqual(self.mix.read_bytes(), mix_before)
+        self.assertEqual(self.catalog.read_bytes(), catalog_before)
+        self.assertEqual(finder_metadata.read_bytes(), b"preserve benign metadata")
+        self.assertFalse(legacy_proposal.exists())
+
+        self._write_inventory()
+        inventory_summary = json.loads((self.build / "01-plan/inventory/inventory-summary.json").read_text())
+        self.assertEqual(inventory_summary["source_count"], 3)
+        self.assertEqual(inventory_summary["source_family_count"], 3)
+        self.assertEqual(inventory_summary["subcategory_count"], 3)
+        self.assertEqual(inventory_summary["category_count"], 4)
+        self.assertEqual(inventory_summary["lower_group_count"], 4)
+        self.assertEqual(inventory_summary["source_uint32_values"], 650)
+        self.assertEqual(inventory_summary["token_delta"], 13_999_999_999_350)
+        self.assertGreater(inventory_summary["sampling_ratio"], 1)
+        self.assertIn("upsample", inventory_summary["sampling_rate"])
+        self.assertEqual(inventory_summary["details_artifact"], "inventory-details.json")
+        inventory_details = json.loads((self.build / "01-plan/inventory/inventory-details.json").read_text())
+        self.assertEqual(inventory_details["source_uint32_values"], 650)
+        self.assertEqual(inventory_details["source_family_count"], 3)
+        self.assertEqual(inventory_details["subcategory_count"], 3)
+        catalog_source = next(
+            source for source in inventory_details["sources"] if source["mix_name"] == "catalog-source:topic"
+        )
+        self.assertEqual(catalog_source["source_uint32_values"], 150)
+        self.assertEqual(catalog_source["source_family"], "catalog-source")
+        self.assertEqual(catalog_source["subcategory_name"], "topic")
+        dropped_category = next(
+            category for category in catalog_source["categories"] if category["category_name"] == "dropped"
+        )
+        self.assertFalse(dropped_category["active"])
+        self.assertEqual(dropped_category["source_uint32_values"], 50)
+        self.assertAlmostEqual(dropped_category["source_percent_of_parent"], 100 / 3)
+        self.assertEqual(
+            dropped_category["lower_groups"][0]["lower_group"],
+            "vigintile_0000",
+        )
+        self.assertEqual(
+            dropped_category["lower_groups"][0]["source_percent_of_parent"],
+            100,
+        )
+        refreshed_summary = refresh_inventory_details(self.build)
+        self.assertEqual(refreshed_summary["source_count"], 3)
+        inventory_plot = (self.build / "01-plan/inventory/plots/available-by-mix.svg").read_text()
+        self.assertIn("% ·", inventory_plot)
+        self.assertIn("tokens", inventory_plot)
+        self.assertFalse((self.build / "01-plan/inventory/plot-data/object-size-bins.csv").exists())
+        self.assertFalse((self.build / "01-plan/inventory/plots/object-size-histogram.svg").exists())
+        inventory_report = (self.build / "01-plan/inventory/report.html").read_text()
+        self.assertIn('data-detail="inventory-family-detail-', inventory_report)
+        self.assertIn('class="subcategory-row', inventory_report)
+        self.assertIn('class="subcategory-detail"', inventory_report)
+        self.assertIn("const setAccordionState", inventory_report)
+        self.assertIn("button.getAttribute('aria-expanded') !== 'true'", inventory_report)
+        self.assertIn('class="comparison-bars"', inventory_report)
+        self.assertIn('class="category-metrics"', inventory_report)
+        self.assertNotIn('class="category-counts"', inventory_report)
+        self.assertIn('class="path-stat"', inventory_report)
+        self.assertIn('<span class="path-stat">1 file', inventory_report)
+        self.assertNotIn('<span class="path-stat">1 NPY', inventory_report)
+        self.assertIn('class="path-metric"', inventory_report)
+        self.assertIn('class="path-detail-metrics"', inventory_report)
+        self.assertIn('class="path-detail-uri"', inventory_report)
+        self.assertIn(
+            "s3://ai2-llm/preprocessed/catalog-source/topic/allenai/tokenizer/0000.npy",
+            inventory_report,
+        )
+        self.assertNotIn(
+            "dolma3p5_pool/catalog-source/topic/allenai/tokenizer/*.npy",
+            inventory_report,
+        )
+        self.assertIn('class="mix-metric-label">Sampling', inventory_report)
+        self.assertIn("Source Inventory and Sampling Plan", inventory_report)
+        self.assertIn('class="summary-metrics"', inventory_report)
+        self.assertIn('class="mix-metrics"', inventory_report)
+        self.assertIn("Overall sampling", inventory_report)
+        self.assertNotIn("S3", inventory_report)
+        self.assertNotIn("NPY bytes", inventory_report)
+        self.assertNotIn("(+", inventory_report)
+        self.assertNotIn("(−", inventory_report)
+        self.assertIn("source ", inventory_report)
+        self.assertIn("% of entry", inventory_report)
+        self.assertIn("% of category", inventory_report)
+        self.assertIn("upsample", inventory_report)
+        self.assertIn("downsample", inventory_report)
+        self.assertIn("vigintile_0000", inventory_report)
+        sampling_path = self.build / "01-plan/inventory/plot-data/sampling-by-lower-group.csv"
+        with sampling_path.open() as f:
+            sampling_paths = list(csv.DictReader(f))
+        dropped = next(row for row in sampling_paths if row["lower_group"] == "vigintile_0000")
+        self.assertEqual(dropped["original_uint32_values"], "50")
+        self.assertEqual(dropped["implied_target_uint32_values"], "0")
+        self.assertIn("downsample", dropped["sampling_rate"])
+        with (self.build / "01-plan/inventory/required-objects.csv").open() as f:
+            self.assertNotIn("resolution_route", csv.DictReader(f).fieldnames)
+        propose_configs(
+            argparse.Namespace(
+                build=self.build,
+                destination_root="s3://test-bucket/new-datasets/dolma3p5",
+                local_temp_root=str(self.root / "temp-base"),
+                max_unit_working_bytes=20_000_000_000_000,
+            )
+        )
+        self.assertEqual(
+            {path.name for path in (self.build / "01-plan").iterdir()},
+            {"resolution", "inventory", "execution", "report.html"},
+        )
+        self.assertFalse((self.build / "01-plan/resolution/report.html").exists())
+        self.assertFalse((self.build / "01-plan/inventory/report.html").exists())
+        self.assertFalse((self.build / "01-plan/execution/report.html").exists())
+        self.assertFalse((self.build / "02-inventory").exists())
+        self.assertFalse((self.build / "03-proposal").exists())
+        configs = list((self.build / "01-plan/execution/config").glob("*.yaml"))
+        self.assertEqual(len(configs), 4)
+        launcher_scripts = list((self.build / "01-plan/execution/launcher-scripts").glob("*.sh"))
+        self.assertEqual(len(launcher_scripts), 4)
+        self.assertTrue(all(path.stat().st_mode & 0o100 for path in launcher_scripts))
+        self.assertTrue(
+            all(
+                '"$python_bin" -P -m dolma.tokenizer.reshard' in path.read_text()
+                and "export PYTHONSAFEPATH=1" in path.read_text()
+                and "cd /tmp" in path.read_text()
+                for path in launcher_scripts
+            )
+        )
+        self.assertTrue(
+            all(
+                "dolma3p5-resharding-status" in path.read_text()
+                and "printf 'running" in path.read_text()
+                and "printf 'succeeded" in path.read_text()
+                and "printf 'failed" in path.read_text()
+                for path in launcher_scripts
+            )
+        )
+        self.assertTrue(all("RESHARDING_MANIFEST_SCHEMA_VERSION" in path.read_text() for path in launcher_scripts))
+        for path in launcher_scripts:
+            self.assertEqual(
+                subprocess.run(["bash", "-n", path], check=False).returncode,
+                0,
+            )
+        with (self.build / "01-plan/execution/config-index.csv").open() as f:
+            execution_units = list(csv.DictReader(f))
+        self.assertEqual(
+            [row["unit_id"] for row in execution_units],
+            [f"{index:0{EXECUTION_UNIT_INDEX_WIDTH}d}" for index in range(len(execution_units))],
+        )
+        self.assertTrue(
+            all(
+                int(row["estimated_peak_local_bytes"]) <= int(row["max_unit_working_bytes"])
+                for row in execution_units
+            )
+        )
+        self.assertTrue(
+            all(
+                1 <= int(row["planned_output_shard_count"]) <= 8
+                and int(row["max_num_files"]) == int(row["planned_output_shard_count"])
+                and int(row["average_output_shard_bytes"])
+                == (int(row["output_npy_bytes"]) + int(row["planned_output_shard_count"]) - 1)
+                // int(row["planned_output_shard_count"])
+                for row in execution_units
+            )
+        )
+        self.assertTrue(all(row["worker_instance_type"] == "i4i.32xlarge" for row in execution_units))
+        self.assertTrue(all(int(row["max_workers"]) == 32 for row in execution_units))
+        dataset_layout = json.loads((self.build / "01-plan/execution/dataset-layout.json").read_text())
+        self.assertEqual(dataset_layout["category_count"], 3)
+        self.assertEqual(dataset_layout["execution_unit_count"], 4)
+        self.assertEqual(dataset_layout["schema_version"], 1)
+        self.assertRegex(
+            dataset_layout["dataset_root"],
+            r"^s3://test-bucket/new-datasets/dolma3p5/dolma3p5-14t-[0-9a-f]{12}$",
+        )
+        self.assertEqual(dataset_layout["layout"], "build-scoped-category-output-v1")
+        self.assertEqual(dataset_layout["execution_unit_index_width"], EXECUTION_UNIT_INDEX_WIDTH)
+        self.assertEqual(dataset_layout["execution_unit_id_width"], EXECUTION_UNIT_INDEX_WIDTH)
+        self.assertEqual(
+            dataset_layout["planned_output_shard_count"],
+            sum(int(row["planned_output_shard_count"]) for row in execution_units),
+        )
+        self.assertEqual(
+            dataset_layout["planned_output_file_count"],
+            dataset_layout["planned_output_shard_count"] * 2,
+        )
+        self.assertEqual(dataset_layout["target_output_shard_bytes"], 64 * 1024**3)
+        self.assertEqual(dataset_layout["max_output_shards_per_unit"], 8)
+        self.assertEqual(_validate_execution_layout(self.build), dataset_layout)
+        runtime_requirements = json.loads((self.build / "01-plan/execution/runtime-requirements.json").read_text())
+        self.assertEqual(runtime_requirements["required_resharding_manifest_schema_version"], 2)
+        for config_path in configs:
+            config = yaml.safe_load(config_path.read_text())
+            self.assertFalse(config["allow_existing_destination"])
+            self.assertEqual(config["s5cmd_download_concurrency"], 32)
+            self.assertTrue(config["source_manifests"])
+            parsed = ReshardingConfig.from_file(config_path)
+            self.assertEqual(len(parsed.source_manifests), 1)
+            self.assertTrue(Path(parsed.source_manifests[0].manifest).is_file())
+            destination = config["destination_prefix"]
+            self.assertRegex(
+                destination,
+                r"^s3://test-bucket/new-datasets/dolma3p5/dolma3p5-14t-[0-9a-f]{12}/",
+            )
+            self.assertRegex(destination, rf"/[0-9]{{{EXECUTION_UNIT_INDEX_WIDTH}}}$")
+            self.assertNotIn("/categories/", destination)
+            self.assertNotIn("/unit-", destination)
+        for row in execution_units:
+            expected_suffix = f"/{row['source_layout_prefix']}/{row['destination_index']}"
+            self.assertIn(expected_suffix, row["destination_prefix"])
+        for manifest_path in (self.build / "01-plan/execution/manifests").glob("*.csv"):
+            with manifest_path.open() as handle:
+                fields = csv.DictReader(handle).fieldnames
+            self.assertIn("partial_target_uint32_values", fields)
+            self.assertIn("selection_seed", fields)
+            self.assertIn("selection_algorithm", fields)
+        for plot in (self.build / "01-plan/execution/plots").glob("*.svg"):
+            ElementTree.parse(plot)
+        self.assertFalse(
+            (self.build / "01-plan/execution/plot-data/category-execution-unit-distribution.csv").exists()
+        )
+        self.assertFalse((self.build / "01-plan/execution/plot-data/most-split-categories.csv").exists())
+        self.assertFalse((self.build / "01-plan/execution/plots/execution-units-per-category.svg").exists())
+        self.assertFalse((self.build / "01-plan/execution/plot-data/object-size-bins.csv").exists())
+        self.assertFalse((self.build / "01-plan/execution/plots/object-size-histogram.svg").exists())
+        self.assertFalse((self.build / "01-plan/execution/plot-data/source-vs-target.csv").exists())
+        self.assertFalse((self.build / "01-plan/execution/plots/source-vs-target.svg").exists())
+        proposal_target_plot = (self.build / "01-plan/execution/plots/target-mix.svg").read_text()
+        self.assertIn("50.00% · 7T tokens", proposal_target_plot)
+        self.assertFalse((self.build / "01-plan/execution/plot-data/target-vs-proposed.csv").exists())
+        self.assertFalse((self.build / "01-plan/execution/plots/target-vs-proposed.svg").exists())
+        self.assertFalse((self.build / "01-plan/execution/plot-data/upsampling-pressure.csv").exists())
+        self.assertFalse((self.build / "01-plan/execution/plots/upsampling-pressure.svg").exists())
+        proposal_report = (self.build / "01-plan/report.html").read_text()
+        self.assertIn("Source inventory &amp; sampling", proposal_report)
+        self.assertIn("Materialization execution", proposal_report)
+        self.assertIn('role="tablist"', proposal_report)
+        self.assertIn('id="source-report-document"', proposal_report)
+        self.assertIn('id="execution-report-document"', proposal_report)
+        self.assertIn('<base href="inventory/">', proposal_report)
+        self.assertIn('<base href="execution/">', proposal_report)
+        self.assertIn("Source Inventory and Sampling Plan", proposal_report)
+        self.assertIn("Materialization Execution Proposal", proposal_report)
+        self.assertIn('class="execution-metrics"', proposal_report)
+        self.assertIn("Execution units", proposal_report)
+        self.assertIn("Output shards", proposal_report)
+        self.assertIn("Output files", proposal_report)
+        self.assertNotIn("Units using document selection", proposal_report)
+        self.assertIn("Planned worker fleet", proposal_report)
+        self.assertIn('class="worker-grid"', proposal_report)
+        self.assertIn("Local NVMe", proposal_report)
+        self.assertIn("Median unit", proposal_report)
+        self.assertIn("P95 unit", proposal_report)
+        self.assertIn("Largest unit", proposal_report)
+        self.assertIn("Median output shard", proposal_report)
+        self.assertIn("P95 output shard", proposal_report)
+        self.assertNotIn("concurrent units on the same worker", proposal_report)
+        self.assertEqual(
+            proposal_report.count('class="execution-unit"'),
+            dataset_layout["execution_unit_count"],
+        )
+        self.assertIn('class="disk-breakdown"', proposal_report)
+        self.assertIn("Source download", proposal_report)
+        self.assertIn("Materialized output", proposal_report)
+        self.assertIn("Selection indexes", proposal_report)
+        self.assertIn("Source shard downloads", proposal_report)
+        self.assertIn("Source shards using document selection", proposal_report)
+        self.assertIn("Reshard concurrency", proposal_report)
+        self.assertIn("Output layout", proposal_report)
+        self.assertIn("shards ·", proposal_report)
+        self.assertIn("Only categories with multiple units", proposal_report)
+        self.assertIn('href="config/', proposal_report)
+        self.assertIn('href="manifests/', proposal_report)
+        self.assertIn('href="launcher-scripts/', proposal_report)
+        self.assertIn("/new-datasets/dolma3p5/", proposal_report)
+        execution_report = proposal_report.split('<template id="execution-report-document">', 1)[1]
+        self.assertNotIn("Overall sampling", execution_report)
+        self.assertNotIn("full copies per object", execution_report)
+        self.assertNotIn("NPYs repeated", execution_report)
+        self.assertNotIn("total object uses", execution_report)
+        self.assertNotIn('class="path-use-summary"', execution_report)
+        self.assertNotIn('class="subcategory-detail"', execution_report)
+        with (self.build / "01-plan/execution/category-allocation.csv").open() as f:
+            allocation_rows = list(csv.DictReader(f))
+        catalog_allocation = next(row for row in allocation_rows if row["mix_name"] == "catalog-source:topic")
+        self.assertEqual(catalog_allocation["available_uint32_values"], "100")
+        self.assertEqual(
+            int(catalog_allocation["planned_uint32_values"]) - int(catalog_allocation["available_uint32_values"]),
+            int(catalog_allocation["token_change_from_original"]),
+        )
+        self.assertEqual(
+            catalog_allocation["minimum_repetition"],
+            catalog_allocation["maximum_repetition"],
+        )
+        self.assertGreater(int(catalog_allocation["maximum_repetition"]), 1)
+        self.assertEqual(int(catalog_allocation["target_residual_uint32_values"]), 0)
+        with (self.build / "01-plan/execution/plot-data/proposed-sampling-by-lower-group.csv").open() as f:
+            proposed_paths = list(csv.DictReader(f))
+        dropped_proposal = next(row for row in proposed_paths if row["lower_group"] == "vigintile_0000")
+        self.assertEqual(dropped_proposal["original_uint32_values"], "50")
+        self.assertEqual(dropped_proposal["proposed_uint32_values"], "0")
+        self.assertEqual(dropped_proposal["token_change_from_original"], "-50")
+        self.assertEqual(dropped_proposal["minimum_repetition"], "0")
+        self.assertEqual(dropped_proposal["maximum_repetition"], "0")
+        self.assertEqual(dropped_proposal["dropped_object_count"], "1")
+        self.assertEqual(dropped_proposal["repeated_object_count"], "0")
+        proposal_summary = json.loads((self.build / "01-plan/execution/proposal-summary.json").read_text())
+        self.assertEqual(proposal_summary["source_uint32_values"], 650)
+        self.assertEqual(
+            proposal_summary["token_change_from_source"],
+            proposal_summary["planned_uint32_values"] - 650,
+        )
+        self.assertEqual(proposal_summary["target_residual_uint32_values"], 0)
+        self.assertEqual(
+            proposal_summary["planned_output_shard_count"],
+            dataset_layout["planned_output_shard_count"],
+        )
+        self.assertEqual(
+            proposal_summary["planned_output_file_count"],
+            dataset_layout["planned_output_file_count"],
+        )
+        self.assertEqual(
+            proposal_summary["document_selection_algorithm"],
+            "document_hash_bucket_v1",
+        )
+
+        with (self.build / "01-plan/inventory/normalized-s3-inventory.csv").open() as f:
+            inventory_rows = [row for row in csv.DictReader(f) if row["required"] == "true"]
+        source_objects = [
+            S3Object(
+                bucket=row["bucket"],
+                key=row["key"],
+                size_bytes=int(row["size_bytes"]),
+                etag=row["etag"],
+                last_modified=row["last_modified"],
+            )
+            for row in inventory_rows
+        ]
+        client = MagicMock()
+        client.list_objects_v2.return_value = {}
+        session = MagicMock()
+        session.client.return_value = client
+        with (
+            patch(
+                "scripts.dolma3p5_resharding.workflow.boto3.Session",
+                return_value=session,
+            ),
+            patch(
+                "scripts.dolma3p5_resharding.workflow._list_prefix",
+                return_value=source_objects,
+            ),
+        ):
+            preflight_build(
+                argparse.Namespace(
+                    build=self.build,
+                    profile=None,
+                    region=None,
+                    max_workers=2,
+                )
+            )
+        self.assertFalse((self.build / "02-preflight/plots").exists())
+        self.assertFalse((self.build / "02-preflight/plot-data").exists())
+        self.assertFalse((self.build / "02-preflight/report.html").exists())
+
+        with (self.build / "01-plan/execution/config-index.csv").open() as f:
+            config_index = list(csv.DictReader(f))
+        output_by_prefix = {}
+        for row in config_index:
+            destination = row["destination_prefix"].removeprefix("s3://")
+            bucket, prefix = destination.split("/", 1)
+            prefix = prefix.rstrip("/") + "/"
+            boundary_residual = (
+                1
+                if int(row["partial_object_count"]) > 0
+                and int(row["allowed_materialized_target_residual_uint32_values"]) >= 1
+                else 0
+            )
+            planned_values = int(row["planned_uint32_values"]) + boundary_residual
+            shard_count = int(row["planned_output_shard_count"])
+            base_values, extra_values = divmod(planned_values, shard_count)
+            output_by_prefix[(bucket, prefix)] = [
+                obj
+                for shard_index in range(shard_count)
+                for obj in (
+                    S3Object(
+                        bucket,
+                        prefix + f"{shard_index:06d}.npy",
+                        (base_values + (shard_index < extra_values)) * 4,
+                        "output-etag",
+                    ),
+                    S3Object(
+                        bucket,
+                        prefix + f"{shard_index:06d}.csv.gz",
+                        24,
+                        "metadata-etag",
+                    ),
+                )
+            ]
+
+        def output_listing(_client, bucket, prefix):
+            return output_by_prefix[(bucket, prefix)]
+
+        with (
+            patch(
+                "scripts.dolma3p5_resharding.workflow.boto3.Session",
+                return_value=session,
+            ),
+            patch(
+                "scripts.dolma3p5_resharding.workflow._list_prefix",
+                side_effect=output_listing,
+            ),
+        ):
+            verify_output(
+                argparse.Namespace(
+                    build=self.build,
+                    profile=None,
+                    region=None,
+                    max_workers=2,
+                )
+            )
+        validate_build(argparse.Namespace(build=self.build))
+        output_summary = json.loads((self.build / "03-output-validation/output-summary.json").read_text())
+        self.assertTrue(output_summary["aggregate_target_residual_within_bound"])
+        self.assertEqual(
+            output_summary["actual_output_shard_count"],
+            output_summary["planned_output_shard_count"],
+        )
+        self.assertNotEqual(
+            output_summary["actual_uint32_values"],
+            output_summary["predicted_uint32_values"],
+        )
+        hierarchy_path = self.build / "03-output-validation/plot-data/source-target-actual.csv"
+        hierarchy_rows = list(csv.DictReader(hierarchy_path.open()))
+        self.assertTrue(
+            {"source_family", "subcategory", "category", "lower_group"} <= {row["level"] for row in hierarchy_rows}
+        )
+        category_row = next(
+            row for row in hierarchy_rows if row["level"] == "category" and row["leaf_id"] == "000:00"
+        )
+        self.assertNotEqual(category_row["actual_uint32_values"], "")
+        lower_group_row = next(
+            row for row in hierarchy_rows if row["level"] == "lower_group" and row["leaf_id"] == "000:00"
+        )
+        self.assertEqual(lower_group_row["actual_uint32_values"], "")
+        output_report = (self.build / "03-output-validation/report.html").read_text()
+        self.assertIn("Source, Target, and Materialized Output", output_report)
+        self.assertIn("Actual − target", output_report)
+        self.assertIn("per-input provenance", output_report)
+        self.assertNotIn("predicted-vs-actual.svg", output_report)
+        propose_configs(
+            argparse.Namespace(
+                build=self.build,
+                destination_root="s3://test-bucket/new-datasets/dolma3p5",
+                local_temp_root=str(self.root / "temp-base"),
+                max_unit_working_bytes=20_000_000_000_000,
+            )
+        )
+        self.assertTrue((self.build / "01-plan/execution/proposal-summary.json").is_file())
+        self.assertFalse((self.build / "02-preflight").exists())
+        self.assertFalse((self.build / "03-output-validation").exists())
+
+    def test_plan_refuses_to_replace_unknown_files_in_a_build(self):
+        self._plan()
+        critical_file = self.build / "source-tokens.npy"
+        critical_file.write_bytes(b"do not replace")
+
+        with self.assertRaisesRegex(PreparationError, "unknown top-level entries"):
+            self._plan()
+
+        self.assertEqual(critical_file.read_bytes(), b"do not replace")
+
+    def test_allocation_is_deterministic_and_size_based(self):
+        first = _allocate_object_sampling(26, [8, 12, 20])
+        second = _allocate_object_sampling(26, [8, 12, 20])
+        self.assertEqual(first, second)
+        repetitions, partial_targets, planned = first
+        self.assertEqual(
+            planned,
+            sum(
+                size * repeat + partial for size, repeat, partial in zip([8, 12, 20], repetitions, partial_targets)
+            ),
+        )
+        self.assertEqual(planned, 26)
+        self.assertEqual(repetitions, [0, 0, 0])
+        self.assertEqual(partial_targets, [5, 8, 13])
+
+    def test_large_leaf_target_is_distributed_across_every_object(self):
+        sizes = [2_993_314_462, 2_930_303_417, 2_852_281_877, 2_863_478_216]
+        repetitions, partial_targets, planned = _allocate_object_sampling(4_810_877_400, sizes)
+        self.assertEqual(planned, 4_810_877_400)
+        self.assertEqual(repetitions, [0, 0, 0, 0])
+        self.assertTrue(all(value > 0 for value in partial_targets))
+        self.assertEqual(sum(partial_targets), planned)
+
+    def test_fractional_rate_samples_the_same_fraction_from_every_shard(self):
+        sizes = [1_000, 2_000, 3_000, 4_000]
+        downsample_repeats, downsample_partials, downsample_total = _allocate_object_sampling(3_000, sizes)
+        self.assertEqual(downsample_repeats, [0, 0, 0, 0])
+        self.assertEqual(downsample_partials, [300, 600, 900, 1_200])
+        self.assertEqual(downsample_total, 3_000)
+
+        upsample_repeats, upsample_partials, upsample_total = _allocate_object_sampling(13_000, sizes)
+        self.assertEqual(upsample_repeats, [1, 1, 1, 1])
+        self.assertEqual(upsample_partials, [300, 600, 900, 1_200])
+        self.assertEqual(upsample_total, 13_000)
+
+    def test_category_execution_units_respect_working_budget(self):
+        row = {
+            "npy_uri": "s3://bucket/tokens.npy",
+            "metadata_uri": "s3://bucket/tokens.csv.gz",
+            "npy_size_bytes": 400,
+            "metadata_size_bytes": 20,
+            "repeat_count": 7,
+        }
+        units = _partition_object_uses([row], max_unit_working_bytes=1_700)
+        self.assertEqual([unit[0]["repeat_count"] for unit in units], [3, 3, 1])
+        self.assertEqual(sum(unit[0]["repeat_count"] for unit in units), 7)
+        with self.assertRaises(PreparationError):
+            _partition_object_uses([row], max_unit_working_bytes=839)
+
+
+class TestReshardingSafety(unittest.TestCase):
+    def setUp(self):
+        self.temp_dir = tempfile.TemporaryDirectory()
+        self.root = Path(self.temp_dir.name)
+
+    def tearDown(self):
+        self.temp_dir.cleanup()
+
+    def _local_manifest(self) -> Path:
+        pair = self.root / "pair"
+        pair.mkdir()
+        npy = pair / "tokens.npy"
+        metadata = pair / "tokens.csv.gz"
+        npy.write_bytes(b"\x00" * 16)
+        metadata.write_text("0,1,id,src,0\n")
+        manifest = self.root / "manifest.csv"
+        with manifest.open("x", newline="") as f:
+            writer = csv.DictWriter(f, fieldnames=["npy_uri", "metadata_uri", "repeat_count"])
+            writer.writeheader()
+            writer.writerow({"npy_uri": npy, "metadata_uri": metadata, "repeat_count": 2})
+        return manifest
+
+    def test_manifest_preserves_exact_local_pair_and_repetition(self):
+        manifest = ReshardingManifestConfig(self._local_manifest())
+        paths = manifest.take(self.root / "run-input", max_workers=1)
+        self.assertEqual(len(paths), 2)
+        self.assertEqual(paths[0], paths[1])
+        self.assertTrue(Path(paths[0].npy_path).is_file())
+
+    def test_partial_manifest_selects_and_materializes_whole_documents(self):
+        pair = self.root / "partial-pair"
+        pair.mkdir()
+        npy = pair / "tokens.npy"
+        metadata = pair / "tokens.csv.gz"
+        np.arange(100, dtype=np.uint32).tofile(npy)
+        boundaries = [0, 7, 19, 44, 70, 100]
+        with gzip.open(metadata, "wt", encoding="utf-8", newline="") as handle:
+            writer = csv.writer(handle)
+            for index, (start, end) in enumerate(zip(boundaries, boundaries[1:])):
+                writer.writerow([start, end, f"doc-{index}", "source", index])
+
+        manifest_path = self.root / "partial-manifest.csv"
+        with manifest_path.open("x", newline="") as handle:
+            writer = csv.DictWriter(
+                handle,
+                fieldnames=[
+                    "npy_uri",
+                    "metadata_uri",
+                    "repeat_count",
+                    "partial_target_uint32_values",
+                    "selection_seed",
+                    "selection_algorithm",
+                    "npy_size_bytes",
+                    "metadata_size_bytes",
+                ],
+            )
+            writer.writeheader()
+            writer.writerow(
+                {
+                    "npy_uri": npy,
+                    "metadata_uri": metadata,
+                    "repeat_count": 0,
+                    "partial_target_uint32_values": 43,
+                    "selection_seed": 1234,
+                    "selection_algorithm": "document_hash_bucket_v1",
+                    "npy_size_bytes": npy.stat().st_size,
+                    "metadata_size_bytes": metadata.stat().st_size,
+                }
+            )
+
+        manifest = ReshardingManifestConfig(manifest_path)
+        first = manifest.take(self.root / "partial-run-1", max_workers=1)
+        second = manifest.take(self.root / "partial-run-2", max_workers=1)
+        self.assertFalse((pair / "selection.csv.gz").exists())
+        self.assertEqual(len(first), 1)
+        self.assertEqual(
+            first[0].selected_uint32_values,
+            second[0].selected_uint32_values,
+        )
+        self.assertLessEqual(abs(first[0].selected_uint32_values - 43), 30)
+        with gzip.open(first[0].selection_path, "rt", encoding="utf-8") as handle:
+            first_selection = handle.read()
+        with gzip.open(second[0].selection_path, "rt", encoding="utf-8") as handle:
+            second_selection = handle.read()
+        self.assertEqual(first_selection, second_selection)
+
+        output = self.root / "partial-output/000000.npy"
+        progress_updates = []
+        merge_result = merge_group(
+            first,
+            output,
+            np.dtype(np.uint32),
+            progress=progress_updates.append,
+        )
+        materialized = np.fromfile(output, dtype=np.uint32)
+        self.assertEqual(len(materialized), first[0].selected_uint32_values)
+        previous_end = 0
+        with gzip.open(output.with_suffix(".csv.gz"), "rt", encoding="utf-8") as handle:
+            rows = list(csv.reader(handle))
+        for row in rows:
+            self.assertEqual(int(row[0]), previous_end)
+            previous_end = int(row[1])
+        self.assertEqual(previous_end, len(materialized))
+        self.assertEqual(merge_result.document_count, len(rows))
+        self.assertEqual(merge_result.token_copy_operations, len(rows))
+        self.assertEqual(merge_result.output_bytes, output.stat().st_size)
+        self.assertEqual(merge_result.metadata_bytes, output.with_suffix(".csv.gz").stat().st_size)
+        self.assertTrue(progress_updates[-1].complete)
+        self.assertEqual(
+            progress_updates[-1].processed_values,
+            merge_result.output_bytes // np.dtype(np.uint32).itemsize,
+        )
+        with self.assertRaises(FileExistsError):
+            merge_group(first, output, np.dtype(np.uint32))
+
+    def test_partial_manifest_samples_documents_from_every_source_shard(self):
+        manifest_rows = []
+        for shard_index, seed in enumerate((101, 202, 303)):
+            pair = self.root / f"distributed-pair-{shard_index}"
+            pair.mkdir()
+            npy = pair / "tokens.npy"
+            metadata = pair / "tokens.csv.gz"
+            np.arange(
+                shard_index * 100,
+                (shard_index + 1) * 100,
+                dtype=np.uint32,
+            ).tofile(npy)
+            with gzip.open(metadata, "wt", encoding="utf-8", newline="") as handle:
+                writer = csv.writer(handle)
+                for document_index in range(10):
+                    start = document_index * 10
+                    writer.writerow(
+                        [
+                            start,
+                            start + 10,
+                            f"shard-{shard_index}-doc-{document_index}",
+                            f"source-{shard_index}",
+                            document_index,
+                        ]
+                    )
+            manifest_rows.append(
+                {
+                    "npy_uri": npy,
+                    "metadata_uri": metadata,
+                    "repeat_count": 0,
+                    "partial_target_uint32_values": 30,
+                    "selection_seed": seed,
+                    "selection_algorithm": "document_hash_bucket_v1",
+                    "npy_size_bytes": npy.stat().st_size,
+                    "metadata_size_bytes": metadata.stat().st_size,
+                }
+            )
+
+        manifest_path = self.root / "distributed-partial-manifest.csv"
+        with manifest_path.open("x", newline="") as handle:
+            writer = csv.DictWriter(handle, fieldnames=list(manifest_rows[0]))
+            writer.writeheader()
+            writer.writerows(manifest_rows)
+
+        paths = ReshardingManifestConfig(manifest_path).take(self.root / "distributed-partial-run", max_workers=2)
+        self.assertEqual(len(paths), 3)
+        selected_index_sets = []
+        for shard_index, path in enumerate(paths):
+            self.assertEqual(path.selected_uint32_values, 30)
+            with gzip.open(path.selection_path, "rt", encoding="utf-8") as handle:
+                selected_rows = list(csv.reader(handle))
+            self.assertEqual(len(selected_rows), 3)
+            self.assertTrue(all(row[2].startswith(f"shard-{shard_index}-") for row in selected_rows))
+            selected_index_sets.append(frozenset(int(row[4]) for row in selected_rows))
+        self.assertGreater(len(set(selected_index_sets)), 1)
+
+    def test_remote_manifest_refuses_source_drift_before_download(self):
+        manifest_path = self.root / "remote-manifest.csv"
+        with manifest_path.open("x", newline="") as f:
+            writer = csv.DictWriter(
+                f,
+                fieldnames=[
+                    "npy_uri",
+                    "metadata_uri",
+                    "repeat_count",
+                    "npy_size_bytes",
+                    "metadata_size_bytes",
+                    "npy_etag",
+                    "metadata_etag",
+                ],
+            )
+            writer.writeheader()
+            writer.writerow(
+                {
+                    "npy_uri": "s3://source-bucket/path/tokens.npy",
+                    "metadata_uri": "s3://source-bucket/path/tokens.csv.gz",
+                    "repeat_count": 1,
+                    "npy_size_bytes": 16,
+                    "metadata_size_bytes": 12,
+                    "npy_etag": "approved-npy",
+                    "metadata_etag": "approved-metadata",
+                }
+            )
+        client = MagicMock()
+        client.head_object.return_value = {
+            "ContentLength": 20,
+            "ETag": '"changed"',
+        }
+        with patch("dolma.tokenizer.reshard.boto3.client", return_value=client):
+            with self.assertRaisesRegex(RuntimeError, "changed before download"):
+                ReshardingManifestConfig(manifest_path).take(self.root / "remote-input", max_workers=1)
+
+    def test_remote_manifest_streams_s5cmd_with_large_file_concurrency(self):
+        manifest_path = self.root / "remote-manifest.csv"
+        with manifest_path.open("x", newline="") as handle:
+            writer = csv.DictWriter(
+                handle,
+                fieldnames=[
+                    "npy_uri",
+                    "metadata_uri",
+                    "repeat_count",
+                    "npy_size_bytes",
+                    "metadata_size_bytes",
+                ],
+            )
+            writer.writeheader()
+            writer.writerow(
+                {
+                    "npy_uri": "s3://source-bucket/path/tokens.npy",
+                    "metadata_uri": "s3://source-bucket/path/tokens.csv.gz",
+                    "repeat_count": 1,
+                    "npy_size_bytes": 16,
+                    "metadata_size_bytes": 12,
+                }
+            )
+
+        client = MagicMock()
+        client.head_object.side_effect = lambda **request: {
+            "ContentLength": 16 if request["Key"].endswith(".npy") else 12,
+            "ETag": '"source"',
+        }
+
+        def emulate_s5cmd(command: list[str], phase: str) -> None:
+            self.assertEqual(command[:5], ["s5cmd", "--stat", "--numworkers", "3", "run"])
+            self.assertEqual(phase, "source download")
+            command_lines = Path(command[-1]).read_text().splitlines()
+            self.assertEqual(len(command_lines), 2)
+            for command_line, expected_size in zip(command_lines, (16, 12)):
+                arguments = shlex.split(command_line)
+                self.assertEqual(
+                    arguments[:6],
+                    [
+                        "cp",
+                        "--show-progress",
+                        "--raw",
+                        "--no-clobber",
+                        "--concurrency",
+                        "17",
+                    ],
+                )
+                destination = Path(arguments[-1])
+                destination.write_bytes(b"\x00" * expected_size)
+
+        with (
+            patch("dolma.tokenizer.reshard.boto3.client", return_value=client),
+            patch("dolma.tokenizer.reshard._run_s5cmd", side_effect=emulate_s5cmd),
+        ):
+            paths = ReshardingManifestConfig(manifest_path).take(
+                self.root / "remote-input",
+                max_workers=3,
+                s5cmd_concurrency=17,
+            )
+
+        self.assertEqual(len(paths), 1)
+        self.assertEqual(Path(paths[0].npy_path).stat().st_size, 16)
+        self.assertEqual(Path(paths[0].csv_path).stat().st_size, 12)
+
+    def test_s5cmd_carriage_return_progress_is_preserved_as_log_records(self):
+        output = io.StringIO()
+        with redirect_stdout(output):
+            run_s5cmd(
+                [
+                    sys.executable,
+                    "-c",
+                    ("import sys; " "sys.stdout.write('10% complete\\r20% complete\\r'); " "sys.stdout.flush()"),
+                ],
+                "test transfer",
+            )
+
+        self.assertEqual(
+            output.getvalue().splitlines(),
+            ["10% complete", "20% complete"],
+        )
+
+    def test_existing_local_destination_is_refused(self):
+        destination = self.root / "existing"
+        destination.mkdir()
+        self.assertTrue(destination_has_objects(destination))
+        config = ReshardingConfig.from_dict(
+            {
+                "destination_prefix": str(destination),
+                "source_manifests": [{"manifest": str(self._local_manifest())}],
+                "max_num_files": 2,
+            }
+        )
+        with self.assertRaises(FileExistsError):
+            reshard(config)
+
+    def test_cleanup_only_removes_run_owned_child(self):
+        manifest = self._local_manifest()
+        temp_base = self.root / "temp-base"
+        temp_base.mkdir()
+        sentinel = temp_base / "keep-me"
+        sentinel.write_text("sentinel")
+        destination = self.root / "new-output"
+        config = ReshardingConfig.from_dict(
+            {
+                "destination_prefix": str(destination),
+                "source_manifests": [{"manifest": str(manifest)}],
+                "local_tempdir": str(temp_base),
+                "max_num_files": 2,
+            }
+        )
+        with patch("dolma.tokenizer.reshard.merge_all_npys"):
+            reshard(config)
+        self.assertEqual(sentinel.read_text(), "sentinel")
+        self.assertEqual([path.name for path in temp_base.iterdir()], ["keep-me"])
+
+    def test_upload_always_uses_no_clobber(self):
+        with patch("dolma.tokenizer.reshard._run_s5cmd") as run:
+            upload_to_s3(self.root, "s3://test-bucket/new/prefix", max_workers=3)
+        command = run.call_args.args[0]
+        self.assertIn("--no-clobber", command)
+        self.assertIn("--show-progress", command)
+        self.assertEqual(command[:5], ["s5cmd", "--stat", "--numworkers", "3", "cp"])
+
+    def test_s3_bucket_root_is_refused_before_listing(self):
+        with self.assertRaises(ValueError):
+            destination_has_objects("s3://test-bucket")
+        client = MagicMock()
+        client.list_objects_v2.return_value = {"Contents": [{"Key": "existing"}]}
+        with patch("dolma.tokenizer.reshard.boto3.client", return_value=client):
+            self.assertTrue(destination_has_objects("s3://test-bucket/new/prefix"))
+
+    def test_worker_storage_setup_is_valid_and_defaults_to_no_action(self):
+        syntax = subprocess.run(
+            ["bash", "-n", str(WORKER_STORAGE_SCRIPT)],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        self.assertEqual(syntax.returncode, 0, syntax.stderr)
+
+        help_result = subprocess.run(
+            ["bash", str(WORKER_STORAGE_SCRIPT), "--help"],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        self.assertEqual(help_result.returncode, 0, help_result.stderr)
+        self.assertIn("--check", help_result.stdout)
+        self.assertIn("--apply", help_result.stdout)
+
+        no_mode = subprocess.run(
+            ["bash", str(WORKER_STORAGE_SCRIPT)],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        self.assertEqual(no_mode.returncode, 2)
+
+    def test_worker_storage_setup_only_selects_instance_store_devices(self):
+        script = WORKER_STORAGE_SCRIPT.read_text()
+        self.assertIn("EC2 NVMe Instance Storage", script)
+        self.assertIn('sudo wipefs -n "$device"', script)
+        self.assertIn('if [[ "$mode" == "--check" ]]', script)
+        self.assertNotIn("/dev/nvme1n1", script)
+
+    def test_worker_storage_check_plans_direct_mount_or_raid0(self):
+        fake_bin = self.root / "fake-bin"
+        fake_bin.mkdir()
+        fake_lsblk = fake_bin / "lsblk"
+        fake_lsblk.write_text("""#!/usr/bin/env bash
+set -euo pipefail
+if [[ "$*" == "-dpno NAME,TYPE" ]]; then
+  echo "/dev/nvme0n1 disk"
+  for ((i = 1; i <= FAKE_NVME_COUNT; i++)); do
+    echo "/dev/nvme${i}n1 disk"
+  done
+elif [[ "$1" == "-dno" && "$2" == "MODEL" ]]; then
+  if [[ "$3" == "/dev/nvme0n1" ]]; then
+    echo "Amazon Elastic Block Store"
+  else
+    echo "Amazon EC2 NVMe Instance Storage"
+  fi
+elif [[ "$1" == "-bdno" && "$2" == "SIZE" ]]; then
+  echo "$FAKE_NVME_SIZE_BYTES"
+elif [[ "$1" == "-nrpo" && "$2" == "NAME,TYPE" ]]; then
+  echo "$3 disk"
+elif [[ "$1" == "-nrpo" && "$2" == "MOUNTPOINT" ]]; then
+  echo
+else
+  echo "unexpected lsblk arguments: $*" >&2
+  exit 97
+fi
+""")
+        fake_sudo = fake_bin / "sudo"
+        fake_sudo.write_text("""#!/usr/bin/env bash
+if [[ "$1" == "wipefs" && "$2" == "-n" ]]; then
+  exit 0
+fi
+echo "unexpected sudo arguments: $*" >&2
+exit 97
+""")
+        fake_findmnt = fake_bin / "findmnt"
+        fake_findmnt.write_text("#!/usr/bin/env bash\nexit 1\n")
+        for command in (fake_lsblk, fake_sudo, fake_findmnt):
+            command.chmod(0o755)
+
+        environment = os.environ.copy()
+        environment["PATH"] = f"{fake_bin}:{environment['PATH']}"
+        for device_count, device_size, layout, expected in (
+            (1, 1_875_000_000_000, "auto", "Plan: format /dev/nvme1n1 as XFS"),
+            (2, 1_875_000_000_000, "auto", "Layout decision required"),
+            (2, 1_875_000_000_000, "single", "Plan: format /dev/nvme1n1 as XFS"),
+            (2, 1_875_000_000_000, "raid0", "create RAID0 across 2 devices"),
+        ):
+            environment["FAKE_NVME_COUNT"] = str(device_count)
+            environment["FAKE_NVME_SIZE_BYTES"] = str(device_size)
+            result = subprocess.run(
+                [
+                    "bash",
+                    str(WORKER_STORAGE_SCRIPT),
+                    "--check",
+                    "--layout",
+                    layout,
+                ],
+                check=False,
+                capture_output=True,
+                text=True,
+                env=environment,
+            )
+            self.assertEqual(result.returncode, 0, result.stderr)
+            self.assertIn(
+                f"Detected {device_count} EC2 instance-store device(s)",
+                result.stdout,
+            )
+            self.assertIn(expected, result.stdout)
+            self.assertIn("no storage changes were made", result.stdout)
+
+        environment["FAKE_NVME_COUNT"] = "2"
+        environment["FAKE_NVME_SIZE_BYTES"] = "1875000000000"
+        implicit_apply = subprocess.run(
+            ["bash", str(WORKER_STORAGE_SCRIPT), "--apply"],
+            check=False,
+            capture_output=True,
+            text=True,
+            env=environment,
+        )
+        self.assertEqual(implicit_apply.returncode, 1)
+        self.assertIn("Refusing to apply an implicit layout", implicit_apply.stderr)
+
+
+class TestDocumentSelection(unittest.TestCase):
+    """Covers create_document_selection directly, including its rejection paths.
+
+    The manifest tests exercise selection end to end.
+    """
+
+    def setUp(self):
+        self.temp_dir = tempfile.TemporaryDirectory()
+        self.root = Path(self.temp_dir.name)
+
+    def tearDown(self):
+        self.temp_dir.cleanup()
+
+    def _write_metadata(self, rows, name="tokens.csv.gz"):
+        path = self.root / name
+        with gzip.open(path, "wt", encoding="utf-8", newline="") as handle:
+            csv.writer(handle).writerows(rows)
+        return path
+
+    def _contiguous_documents(self, lengths, name="tokens.csv.gz"):
+        rows = []
+        start = 0
+        for index, length in enumerate(lengths):
+            rows.append([start, start + length, f"doc-{index}", "source", index])
+            start += length
+        return self._write_metadata(rows, name=name), start
+
+    def _read_selection(self, path):
+        with gzip.open(path, "rt", encoding="utf-8") as handle:
+            return [row for row in csv.reader(handle)]
+
+    def test_same_seed_and_metadata_select_the_same_documents(self):
+        metadata, total = self._contiguous_documents([3 + (i % 7) for i in range(200)])
+        first = create_document_selection(
+            metadata_path=metadata,
+            selection_path=self.root / "first.csv.gz",
+            source_uint32_values=total,
+            target_uint32_values=total // 2,
+            seed=99,
+        )
+        second = create_document_selection(
+            metadata_path=metadata,
+            selection_path=self.root / "second.csv.gz",
+            source_uint32_values=total,
+            target_uint32_values=total // 2,
+            seed=99,
+        )
+        self.assertEqual(
+            self._read_selection(first.selection_path),
+            self._read_selection(second.selection_path),
+        )
+        self.assertEqual(first.selected_uint32_values, second.selected_uint32_values)
+        self.assertEqual(first.selected_document_count, second.selected_document_count)
+
+    def test_a_different_seed_selects_a_different_document_set(self):
+        metadata, total = self._contiguous_documents([3 + (i % 7) for i in range(200)])
+        selections = []
+        for seed in (1, 2):
+            result = create_document_selection(
+                metadata_path=metadata,
+                selection_path=self.root / f"seed-{seed}.csv.gz",
+                source_uint32_values=total,
+                target_uint32_values=total // 2,
+                seed=seed,
+            )
+            selections.append(self._read_selection(result.selection_path))
+        self.assertNotEqual(selections[0], selections[1])
+
+    def test_selection_keeps_whole_documents_within_the_residual_bound(self):
+        lengths = [5 + (i % 11) for i in range(150)]
+        metadata, total = self._contiguous_documents(lengths)
+        target = total // 3
+        result = create_document_selection(
+            metadata_path=metadata,
+            selection_path=self.root / "selection.csv.gz",
+            source_uint32_values=total,
+            target_uint32_values=target,
+            seed=7,
+        )
+        source_rows = {tuple(row) for row in self._read_selection(metadata)}
+        selected_rows = self._read_selection(result.selection_path)
+        for row in selected_rows:
+            self.assertIn(tuple(row), source_rows)
+        self.assertEqual(
+            result.selected_uint32_values,
+            sum(int(row[1]) - int(row[0]) for row in selected_rows),
+        )
+        self.assertEqual(len(selected_rows), result.selected_document_count)
+        self.assertEqual(result.source_document_count, len(lengths))
+        self.assertEqual(result.largest_document_uint32_values, max(lengths))
+        self.assertEqual(
+            result.target_residual_uint32_values,
+            result.selected_uint32_values - target,
+        )
+        self.assertLessEqual(
+            abs(result.target_residual_uint32_values),
+            result.largest_document_uint32_values,
+        )
+
+    def test_progress_reports_both_passes(self):
+        metadata, total = self._contiguous_documents([4] * 25)
+        updates = []
+        create_document_selection(
+            metadata_path=metadata,
+            selection_path=self.root / "selection.csv.gz",
+            source_uint32_values=total,
+            target_uint32_values=total // 2,
+            seed=3,
+            progress=lambda phase, documents, values: updates.append(phase),
+            progress_interval_seconds=0.0,
+        )
+        self.assertIn("pass 1/2", updates)
+        self.assertIn("pass 2/2", updates)
+
+    def test_non_contiguous_metadata_is_rejected(self):
+        metadata = self._write_metadata([[0, 5, "doc-0", "source", 0], [7, 10, "doc-1", "source", 1]])
+        with self.assertRaises(ValueError) as caught:
+            create_document_selection(
+                metadata_path=metadata,
+                selection_path=self.root / "selection.csv.gz",
+                source_uint32_values=10,
+                target_uint32_values=4,
+                seed=1,
+            )
+        self.assertIn("Non-contiguous", str(caught.exception))
+
+    def test_metadata_that_does_not_cover_the_memmap_is_rejected(self):
+        metadata, total = self._contiguous_documents([4, 6])
+        with self.assertRaises(ValueError) as caught:
+            create_document_selection(
+                metadata_path=metadata,
+                selection_path=self.root / "selection.csv.gz",
+                source_uint32_values=total + 10,
+                target_uint32_values=5,
+                seed=1,
+            )
+        self.assertIn("does not cover", str(caught.exception))
+
+    def test_malformed_metadata_row_is_rejected(self):
+        metadata = self._write_metadata([[0, 5, "doc-0", "source"]])
+        with self.assertRaises(ValueError) as caught:
+            create_document_selection(
+                metadata_path=metadata,
+                selection_path=self.root / "selection.csv.gz",
+                source_uint32_values=5,
+                target_uint32_values=2,
+                seed=1,
+            )
+        self.assertIn("expected five columns", str(caught.exception))
+
+    def test_existing_selection_index_is_refused(self):
+        metadata, total = self._contiguous_documents([4] * 5)
+        selection_path = self.root / "selection.csv.gz"
+        selection_path.write_bytes(b"")
+        with self.assertRaises(FileExistsError):
+            create_document_selection(
+                metadata_path=metadata,
+                selection_path=selection_path,
+                source_uint32_values=total,
+                target_uint32_values=total // 2,
+                seed=1,
+            )
+
+    def test_target_outside_the_source_size_is_rejected(self):
+        metadata, total = self._contiguous_documents([4] * 5)
+        for target in (0, total, total + 1):
+            with self.subTest(target=target):
+                with self.assertRaises(ValueError) as caught:
+                    create_document_selection(
+                        metadata_path=metadata,
+                        selection_path=self.root / f"selection-{target}.csv.gz",
+                        source_uint32_values=total,
+                        target_uint32_values=target,
+                        seed=1,
+                    )
+                self.assertIn("target_uint32_values", str(caught.exception))
+
+    def test_algorithm_version_string_is_pinned(self):
+        # Manifests record this string and are rejected on mismatch.
+        self.assertEqual(DOCUMENT_SELECTION_ALGORITHM, "document_hash_bucket_v1")
+
+
+class TestCrossModuleContracts(unittest.TestCase):
+    """Checks constants that workflow.py restates from the library.
+
+    workflow.py stays importable with only boto3 and PyYAML, so it cannot import
+    them from dolma.
+    """
+
+    def test_planner_pins_the_library_document_selection_algorithm(self):
+        self.assertEqual(
+            WORKFLOW_DOCUMENT_SELECTION_ALGORITHM,
+            DOCUMENT_SELECTION_ALGORITHM,
+        )
+
+    def test_planner_pins_the_library_manifest_schema_version(self):
+        self.assertEqual(
+            EXPECTED_MANIFEST_SCHEMA_VERSION,
+            RESHARDING_MANIFEST_SCHEMA_VERSION,
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/tests/python/test_resharding.py b/tests/python/test_resharding.py
index 575ce4ce..e22e052c 100644
--- a/tests/python/test_resharding.py
+++ b/tests/python/test_resharding.py
@@ -11,11 +11,176 @@
 
 from dolma.cli.__main__ import main as cli_main
 from dolma.tokenizer import Tokenizer
-from dolma.tokenizer.reshard import ReshardingConfig, reshard
+from dolma.tokenizer.reshard import (
+    ReshardingConfig,
+    TokensMetadataPaths,
+    group_paths_by_max_num_files,
+    reshard,
+)
+from scripts.resharding.dispatch import (
+    build_poormanray_create_command,
+    build_poormanray_instance_command,
+    build_poormanray_map_command,
+    build_poormanray_run_command,
+)
 
 DOLMA2_TOKENIZER = Path(__file__).parent.parent / "data" / "tokenizer" / "dolma2-test-tokenizer.json"
 
 
+class TestReshardDispatch(unittest.TestCase):
+    def test_repeated_inputs_stay_within_fixed_output_shard_cap(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            npy_path = Path(temp_dir) / "source.npy"
+            npy_path.write_bytes(b"\0" * 4)
+            source = TokensMetadataPaths(str(npy_path), str(npy_path.with_suffix(".csv.gz")))
+
+            groups = group_paths_by_max_num_files([source] * 21, max_num_files=4)
+
+            self.assertEqual(len(groups), 4)
+            self.assertEqual(sum(len(group) for group in groups), 21)
+            self.assertLessEqual(
+                max(len(group) for group in groups) - min(len(group) for group in groups),
+                1,
+            )
+
+    def test_merge_input_log_distinguishes_full_and_selected_views(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            npy_path = Path(temp_dir) / "source.npy"
+            npy_path.write_bytes(b"\0" * 4)
+            metadata_path = npy_path.with_suffix(".csv.gz")
+            full = TokensMetadataPaths(str(npy_path), str(metadata_path))
+            selected = TokensMetadataPaths(
+                str(npy_path),
+                str(metadata_path),
+                selection_path=str(Path(temp_dir) / "selection.csv.gz"),
+                selected_uint32_values=10,
+            )
+
+            with self.assertLogs(level="INFO") as captured:
+                group_paths_by_max_num_files([full, selected], max_num_files=2)
+
+        message = "\n".join(captured.output)
+        self.assertIn(
+            "Merge inputs: 2 uses · 2 distinct inputs · max uses of any input: 1×",
+            message,
+        )
+
+    def test_output_groups_are_balanced_by_bytes(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            paths = []
+            for index, size in enumerate((100, 90, 20, 10)):
+                npy_path = Path(temp_dir) / f"{index:06d}.npy"
+                npy_path.write_bytes(b"\0" * size)
+                paths.append(
+                    TokensMetadataPaths(
+                        str(npy_path),
+                        str(npy_path.with_suffix(".csv.gz")),
+                    )
+                )
+
+            groups = group_paths_by_max_num_files(paths, max_num_files=2)
+
+            self.assertEqual([sum(path.size for path in group) for group in groups], [110, 110])
+
+    def test_poormanray_lifecycle_commands_scope_work_to_selected_instances(self):
+        create = build_poormanray_create_command(
+            cluster="resharding",
+            project="oe-other",
+            region="us-east-1",
+            number=2,
+            instance_type="i4i.2xlarge",
+            storage_type="gp3",
+            storage_size_gib=200,
+            parallelism=2,
+            detach=True,
+        )
+        self.assertEqual(
+            create,
+            [
+                "pmr",
+                "create",
+                "--name",
+                "resharding",
+                "--project",
+                "oe-other",
+                "--region",
+                "us-east-1",
+                "--number",
+                "2",
+                "--instance-type",
+                "i4i.2xlarge",
+                "--storage-type",
+                "gp3",
+                "--storage-size",
+                "200",
+                "--detach",
+                "--parallelism",
+                "2",
+            ],
+        )
+
+        run = build_poormanray_run_command(
+            cluster="resharding",
+            project="oe-other",
+            region="us-east-1",
+            remote_command="true",
+            instance_ids=("i-first", "i-second"),
+            parallelism=2,
+        )
+        self.assertEqual(
+            run[-6:],
+            [
+                "--instance-id",
+                "i-first",
+                "--instance-id",
+                "i-second",
+                "--parallelism",
+                "2",
+            ],
+        )
+
+        resume = build_poormanray_instance_command(
+            "resume",
+            cluster="resharding",
+            project="oe-other",
+            region="us-east-1",
+            instance_ids=("i-first", "i-second"),
+            parallelism=2,
+            detach=True,
+        )
+        self.assertIn("--detach", resume)
+        self.assertEqual(resume.count("--instance-id"), 2)
+
+        command = build_poormanray_map_command(
+            cluster="resharding",
+            project="oe-other",
+            region="us-east-1",
+            script_dir="/tmp/dispatch",
+            spindown=True,
+            instance_ids=("i-first", "i-second"),
+        )
+        self.assertEqual(
+            command,
+            [
+                "pmr",
+                "map",
+                "--name",
+                "resharding",
+                "--project",
+                "oe-other",
+                "--region",
+                "us-east-1",
+                "--script",
+                "/tmp/dispatch",
+                "--spindown",
+                "--instance-id",
+                "i-first",
+                "--instance-id",
+                "i-second",
+            ],
+        )
+
+
 class Sequence(NamedTuple):
     tokens_ids: list[int]
     path: str