Skip to content

Poetry packaging + release 0.2.0 (numpy 2 / econml 0.16 / dowhy 0.14); RunExperiments harness#340

Closed
ZmeiGorynych wants to merge 12 commits into
py-why:mainfrom
ZmeiGorynych:poetry-packaging-numpy2
Closed

Poetry packaging + release 0.2.0 (numpy 2 / econml 0.16 / dowhy 0.14); RunExperiments harness#340
ZmeiGorynych wants to merge 12 commits into
py-why:mainfrom
ZmeiGorynych:poetry-packaging-numpy2

Conversation

@ZmeiGorynych

Copy link
Copy Markdown
Collaborator

Summary

Packages the numpy 2 / econml 0.16 / dowhy 0.14 upgrade for release and adds the RunExperiments harness.

  • Poetry packaging: migrate setup.py → PEP 621 pyproject.toml with the poetry-core build backend; preserve all deps, the test/ray extras, classifiers, and package-data (causalml LICENSE). Add poetry.lock for reproducible installs.
    • The pip/Makefile/CI flow is unchanged and still works (pip builds the poetry-core project), so no CI changes are needed.
  • Release 0.2.0: bump 0.1.60.2.0 (PyPI is already at 0.1.8; minor bump reflects the numpy 2 / econml 0.16 / dowhy 0.14 / py3.10–3.12 dependency overhaul).
  • RunExperiments harness (notebooks/RunExperiments/runners/): estimator-level trial runners (good_*, meta_*, no_meta_* for RCT/KC/KCKP), scores_by_estimator.py, make_summary_tables.py, latex_utils.py; reworked experiment_plots.py / experiment_runner.py; rename rct_no_meta.pymeta_kckp.py.
  • Gitignore large simulation outputs, .venv/, and build artifacts. Add repo CLAUDE.md documenting the DCO sign-off requirement.

This branch was produced by an honest merge of main (no rebase); the file diff is limited to the changes above.

Testing

Local run against the merged stack (Python 3.12, econml 0.16.0 / dowhy 0.14 / numpy 2.2.6):

poetry install --extras test
poetry run pytest -m "not slow"
# 81 passed, 2 skipped

CI will validate the full matrix (ubuntu + windows × py3.10/3.11/3.12).

🤖 Generated with Claude Code

ZmeiGorynych and others added 11 commits December 19, 2024 17:59
Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Signed-off-by: whimo <whimox@gmail.com>
Signed-off-by: whimo <whimox@gmail.com>
…into main

Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Add the experiment-running harness under notebooks/RunExperiments/runners:
- per-scenario trial entrypoints (good_*/meta_*/no_meta_*/no_meta_to_* for
  RCT/KC/KCKP) and estimator_level_trials.py
- scores_by_estimator.py, make_summary_tables.py and latex_utils.py for
  aggregating results into LaTeX summary tables
- rework experiment_plots.py / experiment_runner.py; rename rct_no_meta.py
  -> meta_kckp.py
- gitignore large simulation outputs (EXPERIMENT_RESULTS_*, RunDatasets,
  runners/plots, runners/tex) and stray GCP keys
- add repo CLAUDE.md documenting the DCO sign-off requirement

Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
Signed-off-by: Egor Kraev <egor.kraev@gmail.com>

# Conflicts:
#	causaltune/optimiser.py
#	causaltune/remote.py
#	causaltune/score/bite.py
#	causaltune/score/scoring.py
#	notebooks/RunExperiments/cluster_config.yaml
#	notebooks/RunExperiments/runners/experiment_plots.py
#	notebooks/RunExperiments/runners/experiment_runner.py
#	notebooks/RunExperiments/runners/kc_no_meta.py
#	notebooks/RunExperiments/runners/kckp_no_meta.py
- Migrate setup.py -> PEP 621 pyproject.toml with poetry-core build backend;
  preserve deps, test/ray extras, classifiers, package-data (causalml LICENSE).
- Add poetry.lock for reproducible installs (pip/Makefile/CI flow still works
  via the poetry-core backend, so no CI changes needed).
- Bump version 0.1.6 -> 0.2.0 for the numpy 2 / econml 0.16 / dowhy 0.14
  upgrade merged from main.
- gitignore .venv/ and build artifacts (build/, dist/, *.egg-info/).

Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
@ZmeiGorynych ZmeiGorynych force-pushed the poetry-packaging-numpy2 branch from b7c9291 to 5b7f02c Compare July 7, 2026 09:29
Add a concurrency group (github.workflow + github.ref, cancel-in-progress)
to the Test, Format and Linter workflows so a new push/force-push cancels
the previous still-queued run instead of double-queuing the runner pool.

Signed-off-by: Egor Kraev <egor.kraev@gmail.com>
@ZmeiGorynych

Copy link
Copy Markdown
Collaborator Author

Superseded by #341 (same-repo PR with identical commits). Closing to avoid double-queuing CI. Note: CI is currently stuck org-wide (jobs queue but never start) — GitHub Actions itself is operational, so this looks like a py-why org Actions billing/settings issue, not this PR.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants