Simplify configs

.github/workflows/test.yml

@@ -101,39 +101,38 @@ jobs:
             -W ignore::UserWarning:gymnasium.utils.env_checker:247 \
             && rm -rf tests/outputs outputs
 
-  # TODO(aliberts, rcadene): redesign after v2 migration / removing hydra
-  # end-to-end:
-  #   name: End-to-end
-  #   runs-on: ubuntu-latest
-  #   env:
-  #     MUJOCO_GL: egl
-  #   steps:
-  #     - uses: actions/checkout@v4
-  #       with:
-  #         lfs: true  # Ensure LFS files are pulled
-
-  #     - name: Install apt dependencies
-  #     # portaudio19-dev is needed to install pyaudio
-  #       run: |
-  #         sudo apt-get update && \
-  #         sudo apt-get install -y libegl1-mesa-dev portaudio19-dev
-
-  #     - name: Install poetry
-  #       run: |
-  #         pipx install poetry && poetry config virtualenvs.in-project true
-  #         echo "${{ github.workspace }}/.venv/bin" >> $GITHUB_PATH
-
-  #     - name: Set up Python 3.10
-  #       uses: actions/setup-python@v5
-  #       with:
-  #         python-version: "3.10"
-  #         cache: "poetry"
-
-  #     - name: Install poetry dependencies
-  #       run: |
-  #         poetry install --all-extras
-
-  #     - name: Test end-to-end
-  #       run: |
-  #         make test-end-to-end \
-  #           && rm -rf outputs
+  end-to-end:
+    name: End-to-end
+    runs-on: ubuntu-latest
+    env:
+      MUJOCO_GL: egl
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          lfs: true  # Ensure LFS files are pulled
+
+      - name: Install apt dependencies
+      # portaudio19-dev is needed to install pyaudio
+        run: |
+          sudo apt-get update && \
+          sudo apt-get install -y libegl1-mesa-dev portaudio19-dev
+
+      - name: Install poetry
+        run: |
+          pipx install poetry && poetry config virtualenvs.in-project true
+          echo "${{ github.workspace }}/.venv/bin" >> $GITHUB_PATH
+
+      - name: Set up Python 3.10
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+          cache: "poetry"
+
+      - name: Install poetry dependencies
+        run: |
+          poetry install --all-extras
+
+      - name: Test end-to-end
+        run: |
+          make test-end-to-end \
+            && rm -rf outputs

Makefile

@@ -20,171 +20,159 @@ build-gpu:
 
 test-end-to-end:
 	${MAKE} DEVICE=$(DEVICE) test-act-ete-train
+	${MAKE} DEVICE=$(DEVICE) test-act-ete-train-resume
 	${MAKE} DEVICE=$(DEVICE) test-act-ete-eval
-	${MAKE} DEVICE=$(DEVICE) test-act-ete-train-amp
-	${MAKE} DEVICE=$(DEVICE) test-act-ete-eval-amp
 	${MAKE} DEVICE=$(DEVICE) test-diffusion-ete-train
 	${MAKE} DEVICE=$(DEVICE) test-diffusion-ete-eval
 	${MAKE} DEVICE=$(DEVICE) test-tdmpc-ete-train
-	${MAKE} DEVICE=$(DEVICE) test-tdmpc-ete-train-with-online
 	${MAKE} DEVICE=$(DEVICE) test-tdmpc-ete-eval
-	${MAKE} DEVICE=$(DEVICE) test-default-ete-eval
-	${MAKE} DEVICE=$(DEVICE) test-act-pusht-tutorial
+
+# ${MAKE} DEVICE=$(DEVICE) test-default-ete-eval
+# ${MAKE} DEVICE=$(DEVICE) test-tdmpc-ete-train-with-online
+# ${MAKE} DEVICE=$(DEVICE) test-act-pusht-tutorial
 
 test-act-ete-train:
 	python lerobot/scripts/train.py \
-		policy=act \
-		policy.dim_model=64 \
-		env=aloha \
-		wandb.enable=False \
-		training.offline_steps=2 \
-		training.online_steps=0 \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		device=$(DEVICE) \
-		training.save_checkpoint=true \
-		training.save_freq=2 \
-		policy.n_action_steps=20 \
-		policy.chunk_size=20 \
-		training.batch_size=2 \
-		training.image_transforms.enable=true \
-		hydra.run.dir=tests/outputs/act/
+		--policy.type=act \
+		--policy.dim_model=64 \
+		--policy.n_action_steps=20 \
+		--policy.chunk_size=20 \
+		--env.type=aloha \
+		--env.episode_length=5 \
+		--dataset.repo_id=lerobot/aloha_sim_transfer_cube_human \
+		--dataset.image_transforms.enable=true \
+		--dataset.episodes="[0]" \
+		--batch_size=2 \
+		--offline.steps=4 \
+		--online.steps=0 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--save_freq=2 \
+		--save_checkpoint=true \
+		--log_freq=1 \
+		--wandb.enable=false \
+		--device=$(DEVICE) \
+		--output_dir=tests/outputs/act/
+
+test-act-ete-train-resume:
+	python lerobot/scripts/train.py \
+		--config_path=tests/outputs/act/checkpoints/000002/pretrained_model/train_config.json \
+		--resume=true
 
 test-act-ete-eval:
 	python lerobot/scripts/eval.py \
-		-p tests/outputs/act/checkpoints/000002/pretrained_model \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=8 \
-		device=$(DEVICE) \
-
-test-act-ete-train-amp:
-	python lerobot/scripts/train.py \
-		policy=act \
-		policy.dim_model=64 \
-		env=aloha \
-		wandb.enable=False \
-		training.offline_steps=2 \
-		training.online_steps=0 \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		device=$(DEVICE) \
-		training.save_checkpoint=true \
-		training.save_freq=2 \
-		policy.n_action_steps=20 \
-		policy.chunk_size=20 \
-		training.batch_size=2 \
-		hydra.run.dir=tests/outputs/act_amp/ \
-		training.image_transforms.enable=true \
-		use_amp=true
-
-test-act-ete-eval-amp:
-	python lerobot/scripts/eval.py \
-		-p tests/outputs/act_amp/checkpoints/000002/pretrained_model \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=8 \
-		device=$(DEVICE) \
-		use_amp=true
+		--policy.path=tests/outputs/act/checkpoints/000004/pretrained_model \
+		--env.type=aloha \
+		--env.episode_length=5 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--device=$(DEVICE)
 
 test-diffusion-ete-train:
 	python lerobot/scripts/train.py \
-		policy=diffusion \
-		policy.down_dims=\[64,128,256\] \
-		policy.diffusion_step_embed_dim=32 \
-		policy.num_inference_steps=10 \
-		env=pusht \
-		wandb.enable=False \
-		training.offline_steps=2 \
-		training.online_steps=0 \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		device=$(DEVICE) \
-		training.save_checkpoint=true \
-		training.save_freq=2 \
-		training.batch_size=2 \
-		training.image_transforms.enable=true \
-		hydra.run.dir=tests/outputs/diffusion/
+		--policy.type=diffusion \
+		--policy.down_dims='[64,128,256]' \
+		--policy.diffusion_step_embed_dim=32 \
+		--policy.num_inference_steps=10 \
+		--env.type=pusht \
+		--env.episode_length=5 \
+		--dataset.repo_id=lerobot/pusht \
+		--dataset.image_transforms.enable=true \
+		--dataset.episodes="[0]" \
+		--batch_size=2 \
+		--offline.steps=2 \
+		--online.steps=0 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--save_checkpoint=true \
+		--save_freq=2 \
+		--log_freq=1 \
+		--wandb.enable=false \
+		--device=$(DEVICE) \
+		--output_dir=tests/outputs/diffusion/
 
 test-diffusion-ete-eval:
 	python lerobot/scripts/eval.py \
-		-p tests/outputs/diffusion/checkpoints/000002/pretrained_model \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=8 \
-		device=$(DEVICE) \
+		--policy.path=tests/outputs/diffusion/checkpoints/000002/pretrained_model \
+		--env.type=pusht \
+		--env.episode_length=5 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--device=$(DEVICE)
 
 test-tdmpc-ete-train:
 	python lerobot/scripts/train.py \
-		policy=tdmpc \
-		env=xarm \
-		env.task=XarmLift-v0 \
-		dataset_repo_id=lerobot/xarm_lift_medium \
-		wandb.enable=False \
-		training.offline_steps=2 \
-		training.online_steps=0 \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=2 \
-		device=$(DEVICE) \
-		training.save_checkpoint=true \
-		training.save_freq=2 \
-		training.batch_size=2 \
-		training.image_transforms.enable=true \
-		hydra.run.dir=tests/outputs/tdmpc/
-
-test-tdmpc-ete-train-with-online:
-	python lerobot/scripts/train.py \
-		env=pusht \
-		env.gym.obs_type=environment_state_agent_pos \
-		policy=tdmpc_pusht_keypoints \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=10 \
-		device=$(DEVICE) \
-		training.offline_steps=2 \
-		training.online_steps=20 \
-		training.save_checkpoint=false \
-		training.save_freq=10 \
-		training.batch_size=2 \
-		training.online_rollout_n_episodes=2 \
-		training.online_rollout_batch_size=2 \
-		training.online_steps_between_rollouts=10 \
-		training.online_buffer_capacity=15 \
-		eval.use_async_envs=true \
-		hydra.run.dir=tests/outputs/tdmpc_online/
-
+		--policy.type=tdmpc \
+		--env.type=xarm \
+		--env.task=XarmLift-v0 \
+		--env.episode_length=5 \
+		--dataset.repo_id=lerobot/xarm_lift_medium \
+		--dataset.image_transforms.enable=true \
+		--dataset.episodes='[0]' \
+		--batch_size=2 \
+		--offline.steps=2 \
+		--online.steps=0 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--save_checkpoint=true \
+		--save_freq=2 \
+		--log_freq=1 \
+		--wandb.enable=false \
+		--device=$(DEVICE) \
+		--output_dir=tests/outputs/tdmpc/
 
 test-tdmpc-ete-eval:
 	python lerobot/scripts/eval.py \
-		-p tests/outputs/tdmpc/checkpoints/000002/pretrained_model \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=8 \
-		device=$(DEVICE) \
-
-test-default-ete-eval:
-	python lerobot/scripts/eval.py \
-		--config lerobot/configs/default.yaml \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=8 \
-		device=$(DEVICE) \
-
-test-act-pusht-tutorial:
-	cp examples/advanced/1_train_act_pusht/act_pusht.yaml lerobot/configs/policy/created_by_Makefile.yaml
-	python lerobot/scripts/train.py \
-		policy=created_by_Makefile.yaml \
-		env=pusht \
-		wandb.enable=False \
-		training.offline_steps=2 \
-		eval.n_episodes=1 \
-		eval.batch_size=1 \
-		env.episode_length=2 \
-		device=$(DEVICE) \
-		training.save_model=true \
-		training.save_freq=2 \
-		training.batch_size=2 \
-		training.image_transforms.enable=true \
-		hydra.run.dir=tests/outputs/act_pusht/
-	rm lerobot/configs/policy/created_by_Makefile.yaml
+		--policy.path=tests/outputs/tdmpc/checkpoints/000002/pretrained_model \
+		--env.type=xarm \
+		--env.episode_length=5 \
+		--env.task=XarmLift-v0 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--device=$(DEVICE)
+
+# FIXME: currently broken
+# test-tdmpc-ete-train-with-online:
+# 	python lerobot/scripts/train.py \
+		--policy.type=tdmpc \
+		--env.type=pusht \
+		--env.obs_type=environment_state_agent_pos \
+		--env.episode_length=10 \
+		--dataset.repo_id=lerobot/pusht_keypoints \
+		--dataset.image_transforms.enable=true \
+		--dataset.episodes='[0]' \
+		--batch_size=2 \
+		--offline.steps=2 \
+		--online.steps=20 \
+		--online.rollout_n_episodes=2 \
+		--online.rollout_batch_size=2 \
+		--online.steps_between_rollouts=10 \
+		--online.buffer_capacity=15 \
+		--online.env_seed=10000 \
+		--save_checkpoint=false \
+		--save_freq=10 \
+		--log_freq=1 \
+		--eval.use_async_envs=true \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--device=$(DEVICE) \
+		--output_dir=tests/outputs/tdmpc_online/
+
+# TODO: do we keep this one?
+# test-act-pusht-tutorial:
+# 	cp examples/advanced/1_train_act_pusht/act_pusht.yaml lerobot/configs/policy/created_by_Makefile.yaml
+# 	python lerobot/scripts/train.py \
+# 		policy=created_by_Makefile.yaml \
+# 		env=pusht \
+# 		wandb.enable=False \
+# 		training.offline_steps=2 \
+# 		eval.n_episodes=1 \
+# 		eval.batch_size=1 \
+# 		env.episode_length=2 \
+# 		device=$(DEVICE) \
+# 		training.save_model=true \
+# 		training.save_freq=2 \
+# 		training.batch_size=2 \
+# 		training.image_transforms.enable=true \
+# 		hydra.run.dir=tests/outputs/act_pusht/
+# 	rm lerobot/configs/policy/created_by_Makefile.yaml