[mujoco] Complete dm_control suite tasks#387
Conversation
![github-code-quality[bot]](https://avatars.githubusercontent.com/u/9919?s=60&v=4){kind=small}
| DmcCheetahPixelGymnasiumEnvPool, | ||
| ) = py_env(_DmcCheetahPixelEnvSpec, _DmcCheetahPixelEnvPool) | ||
| ( | ||
| DmcDogEnvSpec, |
| ) = py_env(_DmcCheetahPixelEnvSpec, _DmcCheetahPixelEnvPool) | ||
| ( | ||
| DmcDogEnvSpec, | ||
| DmcDogDMEnvPool, |
| ( | ||
| DmcDogEnvSpec, | ||
| DmcDogDMEnvPool, | ||
| DmcDogGymEnvPool, |
| DmcDogEnvSpec, | ||
| DmcDogDMEnvPool, | ||
| DmcDogGymEnvPool, | ||
| DmcDogGymnasiumEnvPool, |
| DmcDogGymnasiumEnvPool, | ||
| ) = py_env(_DmcDogEnvSpec, _DmcDogEnvPool) | ||
| ( | ||
| DmcDogPixelEnvSpec, |
| ) = py_env(_DmcDogEnvSpec, _DmcDogEnvPool) | ||
| ( | ||
| DmcDogPixelEnvSpec, | ||
| DmcDogPixelDMEnvPool, |
| ( | ||
| DmcDogPixelEnvSpec, | ||
| DmcDogPixelDMEnvPool, | ||
| DmcDogPixelGymEnvPool, |
| DmcDogPixelEnvSpec, | ||
| DmcDogPixelDMEnvPool, | ||
| DmcDogPixelGymEnvPool, | ||
| DmcDogPixelGymnasiumEnvPool, |
| DmcHumanoidCMUPixelGymnasiumEnvPool, | ||
| ) = py_env(_DmcHumanoidCMUPixelEnvSpec, _DmcHumanoidCMUPixelEnvPool) | ||
| ( | ||
| DmcLqrEnvSpec, |
| ) = py_env(_DmcHumanoidCMUPixelEnvSpec, _DmcHumanoidCMUPixelEnvPool) | ||
| ( | ||
| DmcLqrEnvSpec, | ||
| DmcLqrDMEnvPool, |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b4d8ebc7ce
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| mj_addBufferVFS(vfs.get(), asset_name.c_str(), content.data(), | ||
| content.size()); | ||
| } | ||
| AddDirectoryAssetsToVFS(vfs.get(), base_path, "dog_assets"); |
There was a problem hiding this comment.
This call makes every MujocoEnv instance (including non-dog tasks) recursively read and register the entire dog_assets directory before model load. Because MujocoEnv is shared by all DMC domains, environment startup now pays unnecessary filesystem I/O and VFS population cost for hundreds of dog mesh files, which can noticeably slow initialization when creating many envs. Gate this asset loading to dog tasks (or detect references from the XML) so other domains do not regress.
Useful? React with 👍 / 👎.
| mjtNum bump = RandUniform(0.15, 1.0)(gen_); | ||
| model_->hfield_data[start + row * res + col] = bowl * bump; |
There was a problem hiding this comment.
The escape terrain currently multiplies the bowl profile by an independent random value at every heightfield cell, producing high-frequency noise. The dm_control escape task uses a coarse random bump map that is upsampled/smoothed, yielding spatially coherent terrain; without that smoothing, QuadrupedEscape-v1 dynamics and rewards deviate materially from upstream behavior. This should generate coarse bumps and interpolate them rather than sampling each cell independently.
Useful? React with 👍 / 👎.
1 participant
Summary
dm_control==1.0.38, including dog, lqr, quadruped, stacker, and HumanoidCMU walk.suite.ALL_TASKSset fordm_control==1.0.38, including normal and pixel variants for the new domains.This diff completes the dm_control suite task set in EnvPool and tightens alignment coverage for both existing and newly added Mujoco DMC tasks.
Technical Details
envpool/mujoco/dmc/dog.h: dog task initialization, observations, rewards, fetch ball state, and contact sensor handling.envpool/mujoco/dmc/lqr.h: LQR XML generation, action bounds, randomized stiffness alignment, and rewards.envpool/mujoco/dmc/quadruped.h: quadruped terrain/fetch/run/walk logic, including fetch ball body-state alignment.envpool/mujoco/dmc/stacker.h: stacker initialization, target placement, box observations, and box velocity ordering.envpool/mujoco/dmc/utils.cc: XML transforms for dog, lqr, quadruped, and stacker task variants.envpool/mujoco/dmc/mujoco_dmc_suite_ext_align_test.py: alignment coverage for the newly added domains, with per-key dog contact tolerances only forfoot_forces/touch_sensors.envpool/mujoco/dmc/mujoco_dmc_suite_align_test.py: removes an unused pre-step action sample so align tests do not advance NumPy RNG without using the action.foot_forces/touch_sensorsdiverge at roughly1e-4to1e-3absolute scale despite synced reset state; other dog/ext observations stay on the stricterrtol=1e-7path.Test Plan
Automated
bazel test --config=macos --config=test //envpool/mujoco:mujoco_dmc_suite_ext_align_test --test_output=errors: passed.bazel test --config=macos --config=test //envpool/mujoco:mujoco_dmc_suite_align_test --test_output=errors: passed.bazel test --config=macos --config=test //envpool/mujoco:mujoco_dmc_suite_ext_deterministic_test //envpool/mujoco:mujoco_dmc_render_test //envpool/mujoco:mujoco_dmc_pixel_observation_test --test_output=errors: passed.bazel test --config=macos --config=test //envpool/mujoco:mujoco_dmc_suite_deterministic_test --test_output=errors: passed.python -m compileall -q envpool/mujoco/dmc: passed.git diff --check: passed.Suggested Manual
bazel test --config=macos --config=test //envpool:make_test --test_filter=_MakeTest.test_make_mujoco_dmc: run in an environment withcmakeavailable; local attempt was blocked while building the unrelated SDL2/VizDoom dependency becausecmakewas missing.