refactor:(issue #513) rename task g1_sac_wbt → g1_motion_tracking and remove mujo…

README.md

@@ -91,7 +91,7 @@ uv run train --algo sac --task g1_walk_flat --sim motrix
 ```
 
 ```bash
-uv run train --algo sac --task g1_sac_wbt --sim mujoco
+uv run train --algo sac --task g1_motion_tracking --sim mujoco
 ```
 
 ```bash

conf/offpolicy/task/sac/g1_sac_wbt/motrix.yaml → conf/offpolicy/task/sac/g1_motion_tracking/motrix.yaml

@@ -4,7 +4,7 @@
 # backend so checkpoints trained on mujoco can be replayed via motrix's native
 # renderer (`eval --sim motrix`). Training on motrix is not the intended path.
 defaults:
-  - /task/sac/g1_sac_wbt/mujoco
+  - /task/sac/g1_motion_tracking/mujoco
   - _self_
 
 training:

conf/offpolicy/task/sac/g1_wbt_obs/mujoco.yaml

@@ -2,8 +2,7 @@
 # G1 Whole-Body Tracking (WBT) FastSAC on MuJoCo — task ``g1_wbt_obs``.
 #
 # Field-by-field port of the active deploy training profile from
-# ``feat/g1_deploy_v2`` (``conf/offpolicy/task/sac/g1_sac_wbt/mujoco_deploy.yaml``
-# in that branch — Stage 4+5, the one actually used for ONNX-on-real-G1
+# ``feat/g1_deploy_v2`` (Stage 4+5, the one actually used for ONNX-on-real-G1
 # runs; ``mujoco_deploy_v2.yaml`` Plan B was a tuning experiment and is
 # NOT what this profile ports).
 #
@@ -13,7 +12,7 @@
 # The task name binds to ``G1WBTObs`` (see
 # ``src/unilab/envs/motion_tracking/g1/tracking_obs.py``), a strict subclass
 # of ``G1MotionTrackingSACEnv`` — base classes are untouched, so other tasks
-# (``g1_sac_wbt``, ``g1_motion_tracking_deploy``, locomotion / flip / box
+# (``g1_motion_tracking``, ``g1_motion_tracking_deploy``, locomotion / flip / box
 # tracking) are unaffected.
 #
 # Deploy compatibility:

docs/sphinx/source/zh_CN/user_guide/D-tasks/02-g1-motion-tracking.md

@@ -14,15 +14,15 @@
 | 通用全身 tracking | `g1_motion_tracking` | PPO、MLX PPO、APPO | `dance1_subject2_part.npz` |
 | flat flip clip | `g1_flip_tracking` | PPO、MLX PPO、APPO | `flip_360_001__A304.npz` |
 | wall-assisted flip | `g1_wall_flip_tracking` | PPO、MLX PPO、APPO | `flip_from_wall_104__A304.npz` |
-| holosoma-aligned WBT | `g1_sac_wbt` | SAC | 与 `g1_motion_tracking` 共用 |
-| crawl-slope WBT (SAC) | `g1_sac_wbt` + 自定义场景 | SAC | `motion_crawl_slope_uni.npz` |
+| holosoma-aligned WBT | `g1_motion_tracking` | SAC | 与 `g1_motion_tracking` 共用 |
+| crawl-slope WBT (SAC) | `g1_motion_tracking` + 自定义场景 | SAC | `motion_crawl_slope_uni.npz` |
 | sim2real-aligned WBT (deploy chain) | `g1_wbt_obs` | SAC | 与 `g1_motion_tracking` 共用 |
 
 ## 配置入口
 
 - PPO / MLX PPO：`conf/ppo/task/g1_motion_tracking/`、`conf/ppo/task/g1_flip_tracking/`、`conf/ppo/task/g1_wall_flip_tracking/`
 - APPO：`conf/appo/task/g1_motion_tracking/`、`conf/appo/task/g1_flip_tracking/`、`conf/appo/task/g1_wall_flip_tracking/`
-- SAC WBT：`conf/offpolicy/task/sac/g1_sac_wbt/{mujoco,mujoco_deploy,motrix}.yaml`、`conf/offpolicy/task/sac/g1_wbt_obs/mujoco.yaml`
+- SAC WBT：`conf/offpolicy/task/sac/g1_motion_tracking/{mujoco,motrix}.yaml`、`conf/offpolicy/task/sac/g1_wbt_obs/mujoco.yaml`
 
 ## 推荐流程
 
@@ -69,23 +69,22 @@ uv run train --algo ppo --task g1_wall_flip_tracking --sim motrix
 uv run train --algo ppo --task g1_motion_tracking --sim mujoco algo.num_envs=128 algo.max_iterations=5 training.no_play=true
 uv run train --algo appo --task g1_motion_tracking --sim mujoco
 uv run train --algo appo --task g1_motion_tracking --sim motrix
-uv run train --algo sac --task g1_sac_wbt --sim mujoco training.use_amp=true
-uv run train --algo sac --task g1_sac_wbt --sim mujoco --profile deploy training.use_amp=true
+uv run train --algo sac --task g1_motion_tracking --sim mujoco training.use_amp=true
 uv run train --algo sac --task g1_wbt_obs --sim mujoco training.use_amp=true
 uv run eval --algo ppo --task g1_motion_tracking --sim mujoco --load-run -1
 uv run eval --algo ppo --task g1_motion_tracking --sim mujoco --load-run -1 training.cam_tracking=true training.cam_tracking_env_idx=0
-uv run scripts/train_offpolicy.py algo=sac task=sac/g1_sac_wbt/motrix training.play_only=true algo.load_run=/abs/path/to/logs/fast_sac/G1MotionTrackingSAC/2026-04-23_14-06-57_mujoco
+uv run scripts/train_offpolicy.py algo=sac task=sac/g1_motion_tracking/motrix training.play_only=true algo.load_run=/abs/path/to/logs/fast_sac/G1MotionTrackingSAC/2026-04-23_14-06-57_mujoco
 ```
 
-`--profile deploy` 会切到面向部署的独立 SAC owner；`g1_wbt_obs` 是更彻底的 sim2real 部署 task（pelvis IMU + 5 步 per-term 历史 actor obs，与 deploy `ObservationManager` byte-对齐；额外 encoder-bias / foot-friction DR、`joint_acc_l2` / `joint_torque_l2` 奖励；部署工具在 `scripts/deploy/`，obs 对齐由 `tests/scripts/test_obs_alignment_g1_wbt.py` 三路 cross-check）。Motrix sim2sim 回放时用绝对路径透传 `algo.load_run`，不要塞进 `--load-run`。
+`g1_wbt_obs` 是面向 sim2real 部署的 task（pelvis IMU + 5 步 per-term 历史 actor obs，与 deploy `ObservationManager` byte-对齐；额外 encoder-bias / foot-friction DR、`joint_acc_l2` / `joint_torque_l2` 奖励；部署工具在 `scripts/deploy/`，obs 对齐由 `tests/scripts/test_obs_alignment_g1_wbt.py` 三路 cross-check）。Motrix sim2sim 回放时用绝对路径透传 `algo.load_run`，不要塞进 `--load-run`。
 
 ## SAC WBT on crawl-slope 场景
 
-在斜坡地形上跑 `g1_sac_wbt`，需要同时切换 motion 片段和 MuJoCo 场景文件，并固定 episode 长度、关闭 reset 随机化：
+在斜坡地形上跑 `g1_motion_tracking`，需要同时切换 motion 片段和 MuJoCo 场景文件，并固定 episode 长度、关闭 reset 随机化：
 
 ```bash
 CUDA_VISIBLE_DEVICES=1 uv run scripts/train_offpolicy.py \
-  algo=sac task=sac/g1_sac_wbt/mujoco training.use_amp=true algo.seed=1 \
+  algo=sac task=sac/g1_motion_tracking/mujoco training.use_amp=true algo.seed=1 \
   +env.motion_file=src/unilab/assets/motions/g1/motion_crawl_slope_uni.npz \
   +env.scene.model_file=src/unilab/assets/robots/g1/scene_crawl_slope.xml \
   +env.sampling_mode=start \

docs/sphinx/source/zh_CN/user_guide/E-reference/01-backend-support-matrix.md

@@ -87,7 +87,7 @@ uv run scripts/generate_support_matrix.py --write
 | APPO (torch) | `g1_climb_tracking` (g1 climb tracking) | Tested | Tested |
 | SAC (torch) | `g1_walk_flat` (G1 walk flat) | Tested | Tested |
 | SAC (torch) | `g1_walk_rough` (G1 walk rough) | Tested | Tested |
-| SAC (torch) | `g1_sac_wbt` (g1 sac wbt) | Tested | Tested |
+| SAC (torch) | `g1_motion_tracking` (G1 motion tracking) | Tested | Tested |
 | SAC (torch) | `g1_wbt_obs` (g1 wbt obs) | Tested | Registered |
 | TD3 (torch) | `go1_joystick_flat` (Go1 joystick) | Registered | Tested |
 | TD3 (torch) | `go2_joystick_flat` (Go2 joystick) | Registered | Tested |

scripts/deploy/export_deploy_config.py

@@ -64,9 +64,9 @@
 KEYFRAME_NAME = "stand"
 ROOT_QPOS_DIM = 7  # free joint: xyz + quat(wxyz)
 
-# Default obs history length — matches mujoco_deploy.yaml's
+# Default obs history length — matches g1_wbt_obs/mujoco.yaml's
 # `noise_config.obs_history_length`. Override via --obs-history-length when
-# exporting for other training profiles (e.g. mujoco.yaml uses 1).
+# exporting for other training profiles (e.g. g1_motion_tracking/mujoco.yaml uses 1).
 DEFAULT_OBS_HISTORY_LENGTH = 5
 
 
@@ -164,21 +164,21 @@ def main():
         default=DEFAULT_OBS_HISTORY_LENGTH,
         help="Proprio history length H. Must match training-side "
         "noise_config.obs_history_length. Default 5 = current "
-        "mujoco_deploy.yaml. Set 1 for the legacy 154-d schema.",
+        "g1_wbt_obs/mujoco.yaml. Set 1 for the legacy 154-d schema.",
     )
     ap.add_argument(
         "--enable-zero-anchor-pos",
         action="store_true",
         default=True,
         help="Drop motion_anchor_pos_b from actor obs (mjlab parity). "
-        "Matches mujoco_deploy.yaml's noise_config flag.",
+        "Matches g1_wbt_obs/mujoco.yaml's noise_config flag.",
     )
     ap.add_argument(
         "--enable-zero-linvel",
         action="store_true",
         default=True,
         help="Drop base_lin_vel from actor obs (mjlab parity). "
-        "Matches mujoco_deploy.yaml's noise_config flag.",
+        "Matches g1_wbt_obs/mujoco.yaml's noise_config flag.",
     )
     args = ap.parse_args()
 
@@ -253,7 +253,7 @@ def main():
         #   command_joint_pos(29*1) + command_joint_vel(29*1)
         #   + motion_anchor_ori_b(6*1) + gyro(3*5)
         #   + joint_pos_rel(29*5) + dof_vel(29*5) + last_actions(29*5) = 514
-        # Aligns with deploy-profile training run (mujoco_deploy.yaml) which
+        # Aligns with deploy-profile training run (g1_wbt_obs/mujoco.yaml) which
         # drops motion_anchor_pos_b and base_lin_vel from actor obs to match
         # Unitree's verified mjlab "No-State-Estimation" deploy yaml and
         # adds BeyondMimic-style proprio history (default H=5).

tests/benchmark/test_replay_buffer_placement_benchmark.py

@@ -54,8 +54,8 @@ def test_replay_shape_packed_width_includes_critic_fields() -> None:
 
 
 def test_wbt_owner_config_is_only_included_when_present() -> None:
-    assert bench._owner_config_exists("sac", "g1_sac_wbt", "mujoco")
-    assert not bench._owner_config_exists("flashsac", "g1_sac_wbt", "mujoco")
+    assert bench._owner_config_exists("sac", "g1_motion_tracking", "mujoco")
+    assert not bench._owner_config_exists("flashsac", "g1_motion_tracking", "mujoco")
 
 
 def test_default_discovery_includes_existing_offpolicy_mujoco_tasks() -> None:
@@ -68,21 +68,21 @@ def test_default_discovery_includes_existing_offpolicy_mujoco_tasks() -> None:
     assert skipped == []
     assert ("sac", "g1_walk_flat") in targets
     assert ("sac", "g1_walk_rough") in targets
-    assert ("sac", "g1_sac_wbt") in targets
+    assert ("sac", "g1_motion_tracking") in targets
     assert ("flashsac", "g1_walk_flat") in targets
     assert ("flashsac", "go2_joystick_flat") in targets
     assert ("td3", "g1_walk_flat") in targets
-    assert ("flashsac", "g1_sac_wbt") not in targets
+    assert ("flashsac", "g1_motion_tracking") not in targets
 
 
 def test_td3_is_an_allowed_benchmark_algo() -> None:
     assert "td3" in bench._parse_algos("sac,flashsac,td3")
 
 
 def test_parse_tasks_deduplicates_ordered_values() -> None:
-    assert bench._parse_tasks("g1_walk_flat,g1_sac_wbt,g1_walk_flat") == [
+    assert bench._parse_tasks("g1_walk_flat,g1_motion_tracking,g1_walk_flat") == [
         "g1_walk_flat",
-        "g1_sac_wbt",
+        "g1_motion_tracking",
     ]