[WIP] Feat/multithreading

.github/workflows/build.yml

@@ -9,14 +9,14 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ['3.10', '3.11', '3.12']
+        python-version: ["3.10", "3.11", "3.12"]
     steps:
       - uses: actions/checkout@v3
       - run: |
-           docker build -f docker/Dockerfile \
-             --build-arg PYTHON_VERSION=${{ matrix.python-version }} \
-             --tag metaworld-docker .
+          docker build -f docker/Dockerfile \
+            --build-arg PYTHON_VERSION=${{ matrix.python-version }} \
+            --tag metaworld-docker .
       - name: Run tests
-        run: docker run metaworld-docker pytest tests/*
+        run: docker run metaworld-docker pytest -n auto tests/*
 #      - name: Run doctests
 #        run: docker run metaworld-docker pytest --doctest-modules metaworld/

TODO.md

@@ -0,0 +1,11 @@
+## DONEs with reason
+* Renamed `max_path_length` to `max_episode_steps` in several places to align with Gymnasium's terminology.
+* * Removed `Meta-World/goal_hidden` and `Meta-World/goal_observable` as one can just use `Meta-World/MT1` with the appropriate `goal_observable` kwarg.
+
+## TODOs
+
+* Scripts
+* tests
+* Docs
+* Evaluation
+* Test and document meta-batch size there might be the same bug as in master with the division

docs/introduction/basic_usage.md

@@ -135,17 +135,17 @@ envs = gym.make_vec('Meta-World/custom-ml-envs', vector_strategy='sync', envs_li
 ## Arguments
 The gym.make command supports multiple arguments:
 
-| Argument | Usage | Values |
-|----------|-------|--------|
-|  seed    | The number to seed the random number generator with | None or int |
-| max_episode_steps | The maximum number of steps per episode | None or int |
-| use_one_hot | Whether the one hot wrapper should be use to add the task ID to the observation | True or False |
-| num_tasks | The number of parametric variations to sample (default:50) | int |
-| terminate_on_success | Whether to terminate the episode during training when the success signal is seen | True or False|
-| vector_strategy | What kind of vector strategy the environments should be wrapped in | 'sync' or 'async' |
-| task_select | How parametric variations should be selected | "random" or "pseudorandom" |
-| reward_function_version | Use the original reward functions from Meta-World or the updated ones | "v1" or "v2" |
-| reward_normalization_method | Apply a reward normalization wrapper | None or 'gymnasium' or 'exponential' |
-| render_mode | The render mode of each environment | None or 'human' or 'rgb_array' or 'depth_array' |
-| camera_name | The Mujoco name of the camera that should be used to render | 'corner' or 'topview' or 'behindGripper' or 'gripperPOV' or 'corner2' or 'corner3' or 'corner4' |
-| camera_id | The Mujoco ID of the camera that should be used to render | int |
+| Argument                    | Usage                                                                            | Values                                                                                          |
+| --------------------------- | -------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------- |
+| seed                        | The number to seed the random number generator with                              | None or int                                                                                     |
+| max_episode_steps           | The maximum number of steps per episode                                          | None or int                                                                                     |
+| use_one_hot                 | Whether the one hot wrapper should be use to add the task ID to the observation  | True or False                                                                                   |
+| num_tasks                   | The number of parametric variations to sample (default:50)                       | int                                                                                             |
+| terminate_on_success        | Whether to terminate the episode during training when the success signal is seen | True or False                                                                                   |
+| vector_strategy             | What kind of vector strategy the environments should be wrapped in               | 'sync' or 'async'                                                                               |
+| task_sampler                | How parametric variations should be selected                                     | "random" or "pseudorandom"                                                                      |
+| reward_function_version     | Use the original reward functions from Meta-World or the updated ones            | "v1" or "v2"                                                                                    |
+| reward_normalization_method | Apply a reward normalization wrapper                                             | None or 'gymnasium' or 'exponential'                                                            |
+| render_mode                 | The render mode of each environment                                              | None or 'human' or 'rgb_array' or 'depth_array'                                                 |
+| camera_name                 | The Mujoco name of the camera that should be used to render                      | 'corner' or 'topview' or 'behindGripper' or 'gripperPOV' or 'corner2' or 'corner3' or 'corner4' |
+| camera_id                   | The Mujoco ID of the camera that should be used to render                        | int                                                                                             |