Small fixes

* Parse draccus subclass overrides when using --policy.path

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Simon Alibert <75076266+aliberts@users.noreply.github.com>

.cache/calibration/aloha_default/left_follower.json

@@ -1,68 +0,0 @@
-{
-    "homing_offset": [
-        2048,
-        3072,
-        3072,
-        -1024,
-        -1024,
-        2048,
-        -2048,
-        2048,
-        -2048
-    ],
-    "drive_mode": [
-        1,
-        1,
-        1,
-        0,
-        0,
-        1,
-        0,
-        1,
-        0
-    ],
-    "start_pos": [
-        2015,
-        3058,
-        3061,
-        1071,
-        1071,
-        2035,
-        2152,
-        2029,
-        2499
-    ],
-    "end_pos": [
-        -1008,
-        -1963,
-        -1966,
-        2141,
-        2143,
-        -971,
-        3043,
-        -1077,
-        3144
-    ],
-    "calib_mode": [
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "LINEAR"
-    ],
-    "motor_names": [
-        "waist",
-        "shoulder",
-        "shoulder_shadow",
-        "elbow",
-        "elbow_shadow",
-        "forearm_roll",
-        "wrist_angle",
-        "wrist_rotate",
-        "gripper"
-    ]
-}

.cache/calibration/aloha_default/left_leader.json

@@ -1,68 +0,0 @@
-{
-    "homing_offset": [
-        2048,
-        3072,
-        3072,
-        -1024,
-        -1024,
-        2048,
-        -2048,
-        2048,
-        -1024
-    ],
-    "drive_mode": [
-        1,
-        1,
-        1,
-        0,
-        0,
-        1,
-        0,
-        1,
-        0
-    ],
-    "start_pos": [
-        2035,
-        3024,
-        3019,
-        979,
-        981,
-        1982,
-        2166,
-        2124,
-        1968
-    ],
-    "end_pos": [
-        -990,
-        -2017,
-        -2015,
-        2078,
-        2076,
-        -1030,
-        3117,
-        -1016,
-        2556
-    ],
-    "calib_mode": [
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "LINEAR"
-    ],
-    "motor_names": [
-        "waist",
-        "shoulder",
-        "shoulder_shadow",
-        "elbow",
-        "elbow_shadow",
-        "forearm_roll",
-        "wrist_angle",
-        "wrist_rotate",
-        "gripper"
-    ]
-}

.cache/calibration/aloha_default/right_follower.json

@@ -1,68 +0,0 @@
-{
-    "homing_offset": [
-        2048,
-        3072,
-        3072,
-        -1024,
-        -1024,
-        2048,
-        -2048,
-        2048,
-        -2048
-    ],
-    "drive_mode": [
-        1,
-        1,
-        1,
-        0,
-        0,
-        1,
-        0,
-        1,
-        0
-    ],
-    "start_pos": [
-        2056,
-        2895,
-        2896,
-        1191,
-        1190,
-        2018,
-        2051,
-        2056,
-        2509
-    ],
-    "end_pos": [
-        -1040,
-        -2004,
-        -2006,
-        2126,
-        2127,
-        -1010,
-        3050,
-        -1117,
-        3143
-    ],
-    "calib_mode": [
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "LINEAR"
-    ],
-    "motor_names": [
-        "waist",
-        "shoulder",
-        "shoulder_shadow",
-        "elbow",
-        "elbow_shadow",
-        "forearm_roll",
-        "wrist_angle",
-        "wrist_rotate",
-        "gripper"
-    ]
-}

.cache/calibration/aloha_default/right_leader.json

@@ -1,68 +0,0 @@
-{
-    "homing_offset": [
-        2048,
-        3072,
-        3072,
-        -1024,
-        -1024,
-        2048,
-        -2048,
-        2048,
-        -2048
-    ],
-    "drive_mode": [
-        1,
-        1,
-        1,
-        0,
-        0,
-        1,
-        0,
-        1,
-        0
-    ],
-    "start_pos": [
-        2068,
-        3034,
-        3030,
-        1038,
-        1041,
-        1991,
-        1948,
-        2090,
-        1985
-    ],
-    "end_pos": [
-        -1025,
-        -2014,
-        -2015,
-        2058,
-        2060,
-        -955,
-        3091,
-        -940,
-        2576
-    ],
-    "calib_mode": [
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "DEGREE",
-        "LINEAR"
-    ],
-    "motor_names": [
-        "waist",
-        "shoulder",
-        "shoulder_shadow",
-        "elbow",
-        "elbow_shadow",
-        "forearm_roll",
-        "wrist_angle",
-        "wrist_rotate",
-        "gripper"
-    ]
-}

.gitattributes

@@ -11,10 +11,11 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 *.memmap filter=lfs diff=lfs merge=lfs -text
 *.stl filter=lfs diff=lfs merge=lfs -text
 *.safetensors filter=lfs diff=lfs merge=lfs -text
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.arrow filter=lfs diff=lfs merge=lfs -text
 *.json !text !filter !merge !diff
+tests/artifacts/cameras/*.png filter=lfs diff=lfs merge=lfs -text
+*.bag filter=lfs diff=lfs merge=lfs -text

.github/PULL_REQUEST_TEMPLATE.md

@@ -24,7 +24,7 @@ Examples:
 pytest -sx tests/test_stuff.py::test_something
 ```
 ```bash
-python lerobot/scripts/train.py --some.option=true
+python -m lerobot.scripts.train --some.option=true
 ```
 
 ## SECTION TO REMOVE BEFORE SUBMITTING YOUR PR

.github/workflows/build-docker-images.yml

@@ -40,24 +40,24 @@ jobs:
           git lfs install
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0
         with:
           cache-binary: false
 
       - name: Check out code
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           lfs: true
           persist-credentials: false
 
       - name: Login to DockerHub
-        uses: docker/login-action@v3
+        uses: docker/login-action@74a5d142397b4f367a81961eba4e8cd7edddf772 # v3.4.0
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_PASSWORD }}
 
       - name: Build and Push CPU
-        uses: docker/build-push-action@v5
+        uses: docker/build-push-action@ca052bb54ab0790a636c9b5f226502c73d547a25 # v5.4.0
         with:
           context: .
           file: ./docker/lerobot-cpu/Dockerfile
@@ -78,24 +78,24 @@ jobs:
           git lfs install
 
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0
         with:
           cache-binary: false
 
       - name: Check out code
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           lfs: true
           persist-credentials: false
 
       - name: Login to DockerHub
-        uses: docker/login-action@v3
+        uses: docker/login-action@74a5d142397b4f367a81961eba4e8cd7edddf772 # v3.4.0
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_PASSWORD }}
 
       - name: Build and Push GPU
-        uses: docker/build-push-action@v5
+        uses: docker/build-push-action@ca052bb54ab0790a636c9b5f226502c73d547a25 # v5.4.0
         with:
           context: .
           file: ./docker/lerobot-gpu/Dockerfile
@@ -110,23 +110,23 @@ jobs:
       group: aws-general-8-plus
     steps:
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0
         with:
           cache-binary: false
 
       - name: Check out code
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           persist-credentials: false
 
       - name: Login to DockerHub
-        uses: docker/login-action@v3
+        uses: docker/login-action@74a5d142397b4f367a81961eba4e8cd7edddf772 # v3.4.0
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_PASSWORD }}
 
       - name: Build and Push GPU dev
-        uses: docker/build-push-action@v5
+        uses: docker/build-push-action@ca052bb54ab0790a636c9b5f226502c73d547a25 # v5.4.0
         with:
           context: .
           file: ./docker/lerobot-gpu-dev/Dockerfile

.github/workflows/nightly-tests.yml

@@ -33,7 +33,7 @@ jobs:
     runs-on:
       group: aws-general-8-plus
     container:
-      image: huggingface/lerobot-cpu:latest
+      image: huggingface/lerobot-cpu:latest  # zizmor: ignore[unpinned-images]
       options: --shm-size "16gb"
       credentials:
         username: ${{ secrets.DOCKERHUB_USERNAME }}
@@ -44,7 +44,7 @@ jobs:
         working-directory: /lerobot
     steps:
       - name: Tests
-        run: pytest -v --cov=./lerobot --disable-warnings tests
+        run: pytest -v --cov=./src/lerobot --disable-warnings tests
 
       - name: Tests end-to-end
         run: make test-end-to-end
@@ -60,7 +60,7 @@ jobs:
       CUDA_VISIBLE_DEVICES: "0"
       TEST_TYPE: "single_gpu"
     container:
-      image: huggingface/lerobot-gpu:latest
+      image: huggingface/lerobot-gpu:latest  # zizmor: ignore[unpinned-images]
       options: --gpus all --shm-size "16gb"
       credentials:
         username: ${{ secrets.DOCKERHUB_USERNAME }}
@@ -74,7 +74,7 @@ jobs:
         run: nvidia-smi
 
       - name: Test
-        run: pytest -v --cov=./lerobot --cov-report=xml --disable-warnings tests
+        run: pytest -v --cov=./src/lerobot --cov-report=xml --disable-warnings tests
       #   TODO(aliberts): Link with HF Codecov account
       # - name: Upload coverage reports to Codecov with GitHub Action
       #   uses: codecov/codecov-action@v4

.github/workflows/quality.yml

@@ -33,12 +33,12 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout Repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           persist-credentials: false
 
       - name: Set up Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@7f4fc3e22c37d6ff65e88745f38bd3157c663f7c # v4.9.1
         with:
           python-version: ${{ env.PYTHON_VERSION }}
 
@@ -64,9 +64,9 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout Repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           persist-credentials: false
 
       - name: typos-action
-        uses: crate-ci/typos@v1.29.10
+        uses: crate-ci/typos@db35ee91e80fbb447f33b0e5fbddb24d2a1a884f # v1.29.10

.github/workflows/test-docker-build.yml

@@ -35,7 +35,7 @@ jobs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
     steps:
       - name: Check out code
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           persist-credentials: false
 
@@ -64,17 +64,17 @@ jobs:
         docker-file: ${{ fromJson(needs.get_changed_files.outputs.matrix) }}
     steps:
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
+        uses: docker/setup-buildx-action@b5ca514318bd6ebac0fb2aedd5d36ec1b5c232a2 # v3.10.0
         with:
           cache-binary: false
 
       - name: Check out code
-        uses: actions/checkout@v4
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           persist-credentials: false
 
       - name: Build Docker image
-        uses: docker/build-push-action@v5
+        uses: docker/build-push-action@ca052bb54ab0790a636c9b5f226502c73d547a25 # v5.4.0
         with:
           file: ${{ matrix.docker-file }}
           context: .

.github/workflows/test.yml

@@ -17,7 +17,7 @@ name: Tests
 on:
   pull_request:
     paths:
-      - "lerobot/**"
+      - "src/**"
       - "tests/**"
       - "examples/**"
       - ".github/**"
@@ -29,7 +29,7 @@ on:
     branches:
       - main
     paths:
-      - "lerobot/**"
+      - "src/**"
       - "tests/**"
       - "examples/**"
       - ".github/**"
@@ -50,7 +50,7 @@ jobs:
     env:
       MUJOCO_GL: egl
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           lfs: true  # Ensure LFS files are pulled
           persist-credentials: false
@@ -62,7 +62,7 @@ jobs:
           sudo apt-get install -y libegl1-mesa-dev ffmpeg portaudio19-dev
 
       - name: Install uv and python
-        uses: astral-sh/setup-uv@v5
+        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
         with:
           enable-cache: true
           version: ${{ env.UV_VERSION }}
@@ -73,7 +73,7 @@ jobs:
 
       - name: Test with pytest
         run: |
-          uv run pytest tests -v --cov=./lerobot --durations=0 \
+          uv run pytest tests -v --cov=./src/lerobot --durations=0 \
             -W ignore::DeprecationWarning:imageio_ffmpeg._utils:7 \
             -W ignore::UserWarning:torch.utils.data.dataloader:558 \
             -W ignore::UserWarning:gymnasium.utils.env_checker:247 \
@@ -85,7 +85,7 @@ jobs:
     env:
       MUJOCO_GL: egl
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           lfs: true  # Ensure LFS files are pulled
           persist-credentials: false
@@ -94,7 +94,7 @@ jobs:
         run: sudo apt-get update && sudo apt-get install -y ffmpeg
 
       - name: Install uv and python
-        uses: astral-sh/setup-uv@v5
+        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
         with:
           enable-cache: true
           version: ${{ env.UV_VERSION }}
@@ -105,7 +105,7 @@ jobs:
 
       - name: Test with pytest
         run: |
-          uv run pytest tests -v --cov=./lerobot --durations=0 \
+          uv run pytest tests -v --cov=./src/lerobot --durations=0 \
             -W ignore::DeprecationWarning:imageio_ffmpeg._utils:7 \
             -W ignore::UserWarning:torch.utils.data.dataloader:558 \
             -W ignore::UserWarning:gymnasium.utils.env_checker:247 \
@@ -117,7 +117,7 @@ jobs:
     env:
       MUJOCO_GL: egl
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
           lfs: true  # Ensure LFS files are pulled
           persist-credentials: false
@@ -129,7 +129,7 @@ jobs:
           sudo apt-get install -y libegl1-mesa-dev ffmpeg portaudio19-dev
 
       - name: Install uv and python
-        uses: astral-sh/setup-uv@v5
+        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
         with:
           enable-cache: true
           version: ${{ env.UV_VERSION }}

.github/workflows/trufflehog.yml

@@ -24,12 +24,12 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - name: Checkout code
-      uses: actions/checkout@v4
+      uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
       with:
         fetch-depth: 0
         persist-credentials: false
 
     - name: Secret Scanning
-      uses: trufflesecurity/trufflehog@main
+      uses: trufflesecurity/trufflehog@90694bf9af66e7536abc5824e7a87246dbf933cb # v3.88.35
       with:
         extra_args: --only-verified

.gitignore

@@ -12,6 +12,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+# Dev scripts
+.dev
+
 # Logging
 logs
 tmp
@@ -26,6 +29,7 @@ outputs
 
 # VS Code
 .vscode
+.devcontainer
 
 # HPC
 nautilus/*.yaml
@@ -91,10 +95,8 @@ coverage.xml
 .hypothesis/
 .pytest_cache/
 
-# Ignore .cache except calibration
+# Ignore .cache
 .cache/*
-!.cache/calibration/
-!.cache/calibration/**
 
 # Translations
 *.mo

.pre-commit-config.yaml

@@ -37,18 +37,18 @@ repos:
       - id: trailing-whitespace
 
   - repo: https://github.com/adhtruong/mirrors-typos
-    rev: v1.31.1
+    rev: v1.34.0
     hooks:
       - id: typos
         args: [--force-exclude]
 
   - repo: https://github.com/asottile/pyupgrade
-    rev: v3.19.1
+    rev: v3.20.0
     hooks:
     -   id: pyupgrade
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.11.5
+    rev: v0.12.3
     hooks:
       - id: ruff
         args: [--fix]
@@ -57,17 +57,17 @@ repos:
 
   ##### Security #####
   - repo: https://github.com/gitleaks/gitleaks
-    rev: v8.24.3
+    rev: v8.27.2
     hooks:
       - id: gitleaks
 
   - repo: https://github.com/woodruffw/zizmor-pre-commit
-    rev: v1.5.2
+    rev: v1.11.0
     hooks:
       - id: zizmor
 
   - repo: https://github.com/PyCQA/bandit
-    rev: 1.8.3
+    rev: 1.8.6
     hooks:
     - id: bandit
       args: ["-c", "pyproject.toml"]

CONTRIBUTING.md

@@ -67,7 +67,7 @@ post it.
 
 ## Adding new policies, datasets or environments
 
-Look at our implementations for [datasets](./lerobot/common/datasets/), [policies](./lerobot/common/policies/),
+Look at our implementations for [datasets](./src/lerobot/datasets/), [policies](./src/lerobot/policies/),
 environments ([aloha](https://github.com/huggingface/gym-aloha),
 [xarm](https://github.com/huggingface/gym-xarm),
 [pusht](https://github.com/huggingface/gym-pusht))
@@ -269,9 +269,6 @@ Follow these steps to start contributing:
    the PR as a draft PR. These are useful to avoid duplicated work, and to differentiate
    it from PRs ready to be merged;
 4. Make sure existing tests pass;
-<!-- 5. Add high-coverage tests. No quality testing = no merge.
-
-See an example of a good PR here: https://github.com/huggingface/lerobot/pull/ -->
 
 ### Tests
 

MANIFEST.in

@@ -0,0 +1,2 @@
+include src/lerobot/templates/lerobot_modelcard_template.md
+include src/lerobot/datasets/card_template.md

Makefile

@@ -40,14 +40,17 @@ test-end-to-end:
 	${MAKE} DEVICE=$(DEVICE) test-diffusion-ete-eval
 	${MAKE} DEVICE=$(DEVICE) test-tdmpc-ete-train
 	${MAKE} DEVICE=$(DEVICE) test-tdmpc-ete-eval
+	${MAKE} DEVICE=$(DEVICE) test-smolvla-ete-train
+	${MAKE} DEVICE=$(DEVICE) test-smolvla-ete-eval
 
 test-act-ete-train:
-	python lerobot/scripts/train.py \
+	python -m lerobot.scripts.train \
 		--policy.type=act \
 		--policy.dim_model=64 \
 		--policy.n_action_steps=20 \
 		--policy.chunk_size=20 \
 		--policy.device=$(DEVICE) \
+		--policy.push_to_hub=false \
 		--env.type=aloha \
 		--env.episode_length=5 \
 		--dataset.repo_id=lerobot/aloha_sim_transfer_cube_human \
@@ -65,12 +68,12 @@ test-act-ete-train:
 		--output_dir=tests/outputs/act/
 
 test-act-ete-train-resume:
-	python lerobot/scripts/train.py \
+	python -m lerobot.scripts.train \
 		--config_path=tests/outputs/act/checkpoints/000002/pretrained_model/train_config.json \
 		--resume=true
 
 test-act-ete-eval:
-	python lerobot/scripts/eval.py \
+	python -m lerobot.scripts.eval \
 		--policy.path=tests/outputs/act/checkpoints/000004/pretrained_model \
 		--policy.device=$(DEVICE) \
 		--env.type=aloha \
@@ -79,12 +82,13 @@ test-act-ete-eval:
 		--eval.batch_size=1
 
 test-diffusion-ete-train:
-	python lerobot/scripts/train.py \
+	python -m lerobot.scripts.train \
 		--policy.type=diffusion \
 		--policy.down_dims='[64,128,256]' \
 		--policy.diffusion_step_embed_dim=32 \
 		--policy.num_inference_steps=10 \
 		--policy.device=$(DEVICE) \
+		--policy.push_to_hub=false \
 		--env.type=pusht \
 		--env.episode_length=5 \
 		--dataset.repo_id=lerobot/pusht \
@@ -102,7 +106,7 @@ test-diffusion-ete-train:
 		--output_dir=tests/outputs/diffusion/
 
 test-diffusion-ete-eval:
-	python lerobot/scripts/eval.py \
+	python -m lerobot.scripts.eval \
 		--policy.path=tests/outputs/diffusion/checkpoints/000002/pretrained_model \
 		--policy.device=$(DEVICE) \
 		--env.type=pusht \
@@ -111,9 +115,10 @@ test-diffusion-ete-eval:
 		--eval.batch_size=1
 
 test-tdmpc-ete-train:
-	python lerobot/scripts/train.py \
+	python -m lerobot.scripts.train \
 		--policy.type=tdmpc \
 		--policy.device=$(DEVICE) \
+		--policy.push_to_hub=false \
 		--env.type=xarm \
 		--env.task=XarmLift-v0 \
 		--env.episode_length=5 \
@@ -132,7 +137,7 @@ test-tdmpc-ete-train:
 		--output_dir=tests/outputs/tdmpc/
 
 test-tdmpc-ete-eval:
-	python lerobot/scripts/eval.py \
+	python -m lerobot.scripts.eval \
 		--policy.path=tests/outputs/tdmpc/checkpoints/000002/pretrained_model \
 		--policy.device=$(DEVICE) \
 		--env.type=xarm \
@@ -140,3 +145,36 @@ test-tdmpc-ete-eval:
 		--env.task=XarmLift-v0 \
 		--eval.n_episodes=1 \
 		--eval.batch_size=1
+
+
+test-smolvla-ete-train:
+	python -m lerobot.scripts.train \
+		--policy.type=smolvla \
+		--policy.n_action_steps=20 \
+		--policy.chunk_size=20 \
+		--policy.device=$(DEVICE) \
+		--policy.push_to_hub=false \
+		--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 \
+		--steps=4 \
+		--eval_freq=2 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1 \
+		--save_freq=2 \
+		--save_checkpoint=true \
+		--log_freq=1 \
+		--wandb.enable=false \
+		--output_dir=tests/outputs/smolvla/
+
+test-smolvla-ete-eval:
+	python -m lerobot.scripts.eval \
+		--policy.path=tests/outputs/smolvla/checkpoints/000004/pretrained_model \
+		--policy.device=$(DEVICE) \
+		--env.type=aloha \
+		--env.episode_length=5 \
+		--eval.n_episodes=1 \
+		--eval.batch_size=1

README.md

@@ -23,7 +23,30 @@
 </div>
 
 <h2 align="center">
-    <p><a href="https://github.com/huggingface/lerobot/blob/main/examples/12_use_so101.md">
+    <p><a href="https://huggingface.co/docs/lerobot/hope_jr">
+        Build Your Own HopeJR Robot!</a></p>
+</h2>
+
+<div align="center">
+  <img
+    src="media/hope_jr/hopejr.png?raw=true"
+    alt="HopeJR robot"
+    title="HopeJR robot"
+    style="width: 60%;"
+  />
+
+  <p><strong>Meet HopeJR – A humanoid robot arm and hand for dexterous manipulation!</strong></p>
+  <p>Control it with exoskeletons and gloves for precise hand movements.</p>
+  <p>Perfect for advanced manipulation tasks! 🤖</p>
+
+  <p><a href="https://huggingface.co/docs/lerobot/hope_jr">
+      See the full HopeJR tutorial here.</a></p>
+</div>
+
+<br/>
+
+<h2 align="center">
+    <p><a href="https://huggingface.co/docs/lerobot/so101">
         Build Your Own SO-101 Robot!</a></p>
 </h2>
 
@@ -48,11 +71,11 @@
   <p>Train it in minutes with a few simple moves on your laptop.</p>
   <p>Then sit back and watch your creation act autonomously! 🤯</p>
 
-  <p><a href="https://github.com/huggingface/lerobot/blob/main/examples/12_use_so101.md">
+  <p><a href="https://huggingface.co/docs/lerobot/so101">
       See the full SO-101 tutorial here.</a></p>
 
   <p>Want to take it to the next level? Make your SO-101 mobile by building LeKiwi!</p>
-  <p>Check out the <a href="https://github.com/huggingface/lerobot/blob/main/examples/11_use_lekiwi.md">LeKiwi tutorial</a> and bring your robot to life on wheels.</p>
+  <p>Check out the <a href="https://huggingface.co/docs/lerobot/lekiwi">LeKiwi tutorial</a> and bring your robot to life on wheels.</p>
 
   <img src="media/lekiwi/kiwi.webp?raw=true" alt="LeKiwi mobile robot" title="LeKiwi mobile robot" width="50%">
 </div>
@@ -90,6 +113,7 @@
 
 ### Acknowledgment
 
+- The LeRobot team 🤗 for building SmolVLA [Paper](https://arxiv.org/abs/2506.01844), [Blog](https://huggingface.co/blog/smolvla).
 - Thanks to Tony Zhao, Zipeng Fu and colleagues for open sourcing ACT policy, ALOHA environments and datasets. Ours are adapted from [ALOHA](https://tonyzhaozh.github.io/aloha) and [Mobile ALOHA](https://mobile-aloha.github.io).
 - Thanks to Cheng Chi, Zhenjia Xu and colleagues for open sourcing Diffusion policy, Pusht environment and datasets, as well as UMI datasets. Ours are adapted from [Diffusion Policy](https://diffusion-policy.cs.columbia.edu) and [UMI Gripper](https://umi-gripper.github.io).
 - Thanks to Nicklas Hansen, Yunhai Feng and colleagues for open sourcing TDMPC policy, Simxarm environments and datasets. Ours are adapted from [TDMPC](https://github.com/nicklashansen/tdmpc) and [FOWM](https://www.yunhaifeng.com/FOWM).
@@ -129,7 +153,7 @@ pip install -e .
 ```
 
 > **NOTE:** If you encounter build errors, you may need to install additional dependencies (`cmake`, `build-essential`, and `ffmpeg libs`). On Linux, run:
-`sudo apt-get install cmake build-essential python3-dev pkg-config libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libswscale-dev libswresample-dev libavfilter-dev pkg-config`. For other systems, see: [Compiling PyAV](https://pyav.org/docs/develop/overview/installation.html#bring-your-own-ffmpeg)
+`sudo apt-get install cmake build-essential python3-dev pkg-config libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libswscale-dev libswresample-dev libavfilter-dev`. For other systems, see: [Compiling PyAV](https://pyav.org/docs/develop/overview/installation.html#bring-your-own-ffmpeg)
 
 For simulations, 🤗 LeRobot comes with gymnasium environments that can be installed as extras:
 - [aloha](https://github.com/huggingface/gym-aloha)
@@ -148,44 +172,20 @@ wandb login
 
 (note: you will also need to enable WandB in the configuration. See below.)
 
-## Walkthrough
-
-```
-.
-├── examples             # contains demonstration examples, start here to learn about LeRobot
-|   └── advanced         # contains even more examples for those who have mastered the basics
-├── lerobot
-|   ├── configs          # contains config classes with all options that you can override in the command line
-|   ├── common           # contains classes and utilities
-|   |   ├── datasets       # various datasets of human demonstrations: aloha, pusht, xarm
-|   |   ├── envs           # various sim environments: aloha, pusht, xarm
-|   |   ├── policies       # various policies: act, diffusion, tdmpc
-|   |   ├── robot_devices  # various real devices: dynamixel motors, opencv cameras, koch robots
-|   |   └── utils          # various utilities
-|   └── scripts          # contains functions to execute via command line
-|       ├── eval.py                 # load policy and evaluate it on an environment
-|       ├── train.py                # train a policy via imitation learning and/or reinforcement learning
-|       ├── control_robot.py        # teleoperate a real robot, record data, run a policy
-|       ├── push_dataset_to_hub.py  # convert your dataset into LeRobot dataset format and upload it to the Hugging Face hub
-|       └── visualize_dataset.py    # load a dataset and render its demonstrations
-├── outputs               # contains results of scripts execution: logs, videos, model checkpoints
-└── tests                 # contains pytest utilities for continuous integration
-```
-
 ### Visualize datasets
 
 Check out [example 1](./examples/1_load_lerobot_dataset.py) that illustrates how to use our dataset class which automatically downloads data from the Hugging Face hub.
 
 You can also locally visualize episodes from a dataset on the hub by executing our script from the command line:
 ```bash
-python lerobot/scripts/visualize_dataset.py \
+python -m lerobot.scripts.visualize_dataset \
     --repo-id lerobot/pusht \
     --episode-index 0
 ```
 
 or from a dataset in a local folder with the `root` option and the `--local-files-only` (in the following case the dataset will be searched for in `./my_local_data_dir/lerobot/pusht`)
 ```bash
-python lerobot/scripts/visualize_dataset.py \
+python -m lerobot.scripts.visualize_dataset \
     --repo-id lerobot/pusht \
     --root ./my_local_data_dir \
     --local-files-only 1 \
@@ -198,7 +198,7 @@ It will open `rerun.io` and display the camera streams, robot states and actions
 https://github-production-user-asset-6210df.s3.amazonaws.com/4681518/328035972-fd46b787-b532-47e2-bb6f-fd536a55a7ed.mov?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAVCODYLSA53PQK4ZA%2F20240505%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20240505T172924Z&X-Amz-Expires=300&X-Amz-Signature=d680b26c532eeaf80740f08af3320d22ad0b8a4e4da1bcc4f33142c15b509eda&X-Amz-SignedHeaders=host&actor_id=24889239&key_id=0&repo_id=748713144
 
 
-Our script can also visualize datasets stored on a distant server. See `python lerobot/scripts/visualize_dataset.py --help` for more instructions.
+Our script can also visualize datasets stored on a distant server. See `python -m lerobot.scripts.visualize_dataset --help` for more instructions.
 
 ### The `LeRobotDataset` format
 
@@ -251,7 +251,7 @@ Check out [example 2](./examples/2_evaluate_pretrained_policy.py) that illustrat
 
 We also provide a more capable script to parallelize the evaluation over multiple environments during the same rollout. Here is an example with a pretrained model hosted on [lerobot/diffusion_pusht](https://huggingface.co/lerobot/diffusion_pusht):
 ```bash
-python lerobot/scripts/eval.py \
+python -m lerobot.scripts.eval \
     --policy.path=lerobot/diffusion_pusht \
     --env.type=pusht \
     --eval.batch_size=10 \
@@ -263,10 +263,10 @@ python lerobot/scripts/eval.py \
 Note: After training your own policy, you can re-evaluate the checkpoints with:
 
 ```bash
-python lerobot/scripts/eval.py --policy.path={OUTPUT_DIR}/checkpoints/last/pretrained_model
+python -m lerobot.scripts.eval --policy.path={OUTPUT_DIR}/checkpoints/last/pretrained_model
 ```
 
-See `python lerobot/scripts/eval.py --help` for more instructions.
+See `python -m lerobot.scripts.eval --help` for more instructions.
 
 ### Train your own policy
 
@@ -278,14 +278,14 @@ A link to the wandb logs for the run will also show up in yellow in your termina
 
 ![](media/wandb.png)
 
-Note: For efficiency, during training every checkpoint is evaluated on a low number of episodes. You may use `--eval.n_episodes=500` to evaluate on more episodes than the default. Or, after training, you may want to re-evaluate your best checkpoints on more episodes or change the evaluation settings. See `python lerobot/scripts/eval.py --help` for more instructions.
+Note: For efficiency, during training every checkpoint is evaluated on a low number of episodes. You may use `--eval.n_episodes=500` to evaluate on more episodes than the default. Or, after training, you may want to re-evaluate your best checkpoints on more episodes or change the evaluation settings. See `python -m lerobot.scripts.eval --help` for more instructions.
 
 #### Reproduce state-of-the-art (SOTA)
 
 We provide some pretrained policies on our [hub page](https://huggingface.co/lerobot) that can achieve state-of-the-art performances.
 You can reproduce their training by loading the config from their run. Simply running:
 ```bash
-python lerobot/scripts/train.py --config_path=lerobot/diffusion_pusht
+python -m lerobot.scripts.train --config_path=lerobot/diffusion_pusht
 ```
 reproduces SOTA results for Diffusion Policy on the PushT task.
 
@@ -311,7 +311,7 @@ python lerobot/scripts/push_dataset_to_hub.py \
 
 See `python lerobot/scripts/push_dataset_to_hub.py --help` for more instructions.
 
-If your dataset format is not supported, implement your own in `lerobot/common/datasets/push_dataset_to_hub/${raw_format}_format.py` by copying examples like [pusht_zarr](https://github.com/huggingface/lerobot/blob/main/lerobot/common/datasets/push_dataset_to_hub/pusht_zarr_format.py), [umi_zarr](https://github.com/huggingface/lerobot/blob/main/lerobot/common/datasets/push_dataset_to_hub/umi_zarr_format.py), [aloha_hdf5](https://github.com/huggingface/lerobot/blob/main/lerobot/common/datasets/push_dataset_to_hub/aloha_hdf5_format.py), or [xarm_pkl](https://github.com/huggingface/lerobot/blob/main/lerobot/common/datasets/push_dataset_to_hub/xarm_pkl_format.py). -->
+If your dataset format is not supported, implement your own in `lerobot/datasets/push_dataset_to_hub/${raw_format}_format.py` by copying examples like [pusht_zarr](https://github.com/huggingface/lerobot/blob/main/lerobot/datasets/push_dataset_to_hub/pusht_zarr_format.py), [umi_zarr](https://github.com/huggingface/lerobot/blob/main/lerobot/datasets/push_dataset_to_hub/umi_zarr_format.py), [aloha_hdf5](https://github.com/huggingface/lerobot/blob/main/lerobot/datasets/push_dataset_to_hub/aloha_hdf5_format.py), or [xarm_pkl](https://github.com/huggingface/lerobot/blob/main/lerobot/datasets/push_dataset_to_hub/xarm_pkl_format.py). -->
 
 
 ### Add a pretrained policy
@@ -360,7 +360,7 @@ with profile(
 If you want, you can cite this work with:
 ```bibtex
 @misc{cadene2024lerobot,
-    author = {Cadene, Remi and Alibert, Simon and Soare, Alexander and Gallouedec, Quentin and Zouitine, Adil and Wolf, Thomas},
+    author = {Cadene, Remi and Alibert, Simon and Soare, Alexander and Gallouedec, Quentin and Zouitine, Adil and Palma, Steven and Kooijmans, Pepijn and Aractingi, Michel and Shukor, Mustafa and Aubakirova, Dana and Russi, Martino and Capuano, Francesco and Pascale, Caroline and Choghari, Jade and Moss, Jess and Wolf, Thomas},
     title = {LeRobot: State-of-the-art Machine Learning for Real-World Robotics in Pytorch},
     howpublished = "\url{https://github.com/huggingface/lerobot}",
     year = {2024}
@@ -368,6 +368,15 @@ If you want, you can cite this work with:
 ```
 
 Additionally, if you are using any of the particular policy architecture, pretrained models, or datasets, it is recommended to cite the original authors of the work as they appear below:
+- [SmolVLA](https://arxiv.org/abs/2506.01844)
+```bibtex
+@article{shukor2025smolvla,
+  title={SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics},
+  author={Shukor, Mustafa and Aubakirova, Dana and Capuano, Francesco and Kooijmans, Pepijn and Palma, Steven and Zouitine, Adil and Aractingi, Michel and Pascal, Caroline and Russi, Martino and Marafioti, Andres and Alibert, Simon and Cord, Matthieu and Wolf, Thomas and Cadene, Remi},
+  journal={arXiv preprint arXiv:2506.01844},
+  year={2025}
+}
+```
 
 - [Diffusion Policy](https://diffusion-policy.cs.columbia.edu)
 ```bibtex
@@ -408,6 +417,19 @@ Additionally, if you are using any of the particular policy architecture, pretra
   year={2024}
 }
 ```
+
+
+- [HIL-SERL](https://hil-serl.github.io/)
+```bibtex
+@Article{luo2024hilserl,
+title={Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning},
+author={Jianlan Luo and Charles Xu and Jeffrey Wu and Sergey Levine},
+year={2024},
+eprint={2410.21845},
+archivePrefix={arXiv},
+primaryClass={cs.RO}
+}
+```
 ## Star History
 
 [![Star History Chart](https://api.star-history.com/svg?repos=huggingface/lerobot&type=Timeline)](https://star-history.com/#huggingface/lerobot&Timeline)

benchmarks/video/capture_camera_feed.py

@@ -55,7 +55,7 @@ def display_and_save_video_stream(output_dir: Path, fps: int, width: int, height
         if not ret:
             print("Error: Could not read frame.")
             break
-        rr.log("video/stream", rr.Image(frame.numpy()), static=True)
+        rr.log("video/stream", rr.Image(frame), static=True)
         cv2.imwrite(str(capture_dir / f"frame_{frame_index:06d}.png"), frame)
         frame_index += 1
 

benchmarks/video/run_video_benchmark.py

@@ -35,12 +35,12 @@ import torch
 from skimage.metrics import mean_squared_error, peak_signal_noise_ratio, structural_similarity
 from tqdm import tqdm
 
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
-from lerobot.common.datasets.video_utils import (
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.video_utils import (
     decode_video_frames_torchvision,
     encode_video_frames,
 )
-from lerobot.common.utils.benchmark import TimeBenchmark
+from lerobot.utils.benchmark import TimeBenchmark
 
 BASE_ENCODING = OrderedDict(
     [

docker/lerobot-cpu/Dockerfile

@@ -22,7 +22,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 COPY . /lerobot
 WORKDIR /lerobot
 RUN /opt/venv/bin/pip install --upgrade --no-cache-dir pip \
-    && /opt/venv/bin/pip install --no-cache-dir ".[test, aloha, xarm, pusht, dynamixel]" \
+    && /opt/venv/bin/pip install --no-cache-dir ".[test, aloha, xarm, pusht, smolvla]" \
         --extra-index-url https://download.pytorch.org/whl/cpu
 
 # Execute in bash shell rather than python

docker/lerobot-gpu/Dockerfile

@@ -21,4 +21,4 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 COPY . /lerobot
 WORKDIR /lerobot
 RUN /opt/venv/bin/pip install --upgrade --no-cache-dir pip \
-    && /opt/venv/bin/pip install --no-cache-dir ".[test, aloha, xarm, pusht, dynamixel]"
+    && /opt/venv/bin/pip install --no-cache-dir ".[test, aloha, xarm, pusht, dynamixel, smolvla]"

docs/source/_toctree.yml

@@ -5,8 +5,44 @@
     title: Installation
   title: Get started
 - sections:
-  - local: assemble_so101
-    title: Assemble SO-101
-  - local: getting_started_real_world_robot
-    title: Getting Started with Real-World Robots
+  - local: il_robots
+    title: Imitation Learning for Robots
+  - local: il_sim
+    title: Imitation Learning in Sim
+  - local: cameras
+    title: Cameras
+  - local: integrate_hardware
+    title: Bring Your Own Hardware
+  - local: hilserl
+    title: Train a Robot with RL
+  - local: hilserl_sim
+    title: Train RL in Simulation
+  - local: async
+    title: Use Async Inference
   title: "Tutorials"
+- sections:
+  - local: smolvla
+    title: Finetune SmolVLA
+  title: "Policies"
+- sections:
+  - local: hope_jr
+    title: Hope Jr
+  - local: so101
+    title: SO-101
+  - local: so100
+    title: SO-100
+  - local: koch
+    title: Koch v1.1
+  - local: lekiwi
+    title: LeKiwi
+  title: "Robots"
+- sections:
+  - local: notebooks
+    title: Notebooks
+  title: "Resources"
+- sections:
+  - local: contributing
+    title: Contribute to LeRobot
+  - local: backwardcomp
+    title: Backward compatibility
+  title: "About"

docs/source/assemble_so101.mdx

@@ -1,346 +0,0 @@
-# Assemble SO-101
-
-In the steps below we explain how to assemble our flagship robot, the SO-101.
-
-## Source the parts
-
-Follow this [README](https://github.com/TheRobotStudio/SO-ARM100). It contains the bill of materials, with a link to source the parts, as well as the instructions to 3D print the parts,
-and advice if it's your first time printing or if you don't own a 3D printer.
-
-Before assembling, you will first need to configure your motors. To this end, we provide a nice script, so let's first install LeRobot. After configuration, we will also guide you through assembly.
-
-## Install LeRobot
-
-To install LeRobot follow our [Installation Guide](./installation)
-
-## Configure motors
-
-To configure the motors designate one bus servo adapter and 6 motors for your leader arm, and similarly the other bus servo adapter and 6 motors for the follower arm. It's convenient to label them and write on each motor if it's for the follower `F` or for the leader `L` and it's ID from 1 to 6.
-
-You now should plug the 5V or 12V power supply to the motor bus. 5V for the STS3215 7.4V motors and 12V for the STS3215 12V motors. Note that the leader arm always uses the 7.4V motors, so watch out that you plug in the right power supply if you have 12V and 7.4V motors, otherwise you might burn your motors! Now, connect the motor bus to your computer via USB. Note that the USB doesn't provide any power, and both the power supply and USB have to be plugged in.
-
-### Find the USB ports associated to each arm
-
-To find the port for each bus servo adapter, run this script:
-```bash
-python lerobot/scripts/find_motors_bus_port.py
-```
-##### Example outputs of script
-
-<hfoptions id="example">
-<hfoption id="Mac">
-
-Example output leader arm's port: `/dev/tty.usbmodem575E0031751`
-
-```bash
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this MotorsBus is /dev/tty.usbmodem575E0031751
-Reconnect the usb cable.
-```
-
-Example output follower arm port: `/dev/tty.usbmodem575E0032081`
-
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this MotorsBus is /dev/tty.usbmodem575E0032081
-Reconnect the usb cable.
-```
-
-</hfoption>
-<hfoption id="Linux">
-
-On Linux, you might need to give access to the USB ports by running:
-```bash
-sudo chmod 666 /dev/ttyACM0
-sudo chmod 666 /dev/ttyACM1
-```
-
-Example output leader arm port: `/dev/ttyACM0`
-
-```bash
-Finding all available ports for the MotorBus.
-['/dev/ttyACM0', '/dev/ttyACM1']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this MotorsBus is /dev/ttyACM0
-Reconnect the usb cable.
-```
-
-Example output follower arm port: `/dev/ttyACM1`
-
-```
-Finding all available ports for the MotorBus.
-['/dev/ttyACM0', '/dev/ttyACM1']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this MotorsBus is /dev/ttyACM1
-Reconnect the usb cable.
-```
-</hfoption>
-</hfoptions>
-
-#### Update config file
-
-Now that you have your ports, update the **port** default values of [`SO101RobotConfig`](https://github.com/huggingface/lerobot/blob/main/lerobot/common/robot_devices/robots/configs.py).
-You will find something like, update the `port` values with your actual motor ports:
-```python
-@RobotConfig.register_subclass("so101")
-@dataclass
-class So101RobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/so101"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-```
-
-Here is a video of the process:
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/lerobot-find-motorbus.mp4" type="video/mp4" />
-  </video>
- </div>
-
-## Step-by-Step Assembly Instructions
-
-The follower arm uses 6x STS3215 motors with 1/345 gearing. The leader however uses three differently geared motors to make sure it can both sustain its own weight and it can be moved without requiring much force. Which motor is needed for which joint is shown in table below.
-
-| Leader-Arm Axis | Motor | Gear Ratio |
-|-----------------|:-------:|:----------:|
-| Base / Shoulder Yaw | 1 | 1 / 191 |
-| Shoulder Pitch      | 2 | 1 / 345 |
-| Elbow               | 3 | 1 / 191 |
-| Wrist Roll          | 4 | 1 / 147 |
-| Wrist Pitch         | 5 | 1 / 147 |
-| Gripper             | 6 | 1 / 147 |
-
-### Set motor IDs
-
-Plug your motor in one of the two ports of the motor bus and run this script to set its ID to 1. Replace the text after --port to the corresponding control board port.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 1
-```
-
-Then unplug your motor and plug the second motor and set its ID to 2.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 2
-```
-
-Redo this process for all your motors until ID 6. Do the same for the 6 motors of the leader arm, but make sure to change the power supply if you use motors with different voltage and make sure you give the right ID to the right motor according to the table above.
-
-Here is a video of the process:
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/lerobot-configure-motor.mp4" type="video/mp4" />
-  </video>
-</div>
-
-### Clean Parts
-Remove all support material from the 3D-printed parts, the easiest way to do this is using a small screwdriver to get underneath the support material.
-
-### Joint 1
-
-- Place the first motor into the base.
-- Fasten the motor with 4 M2x6mm screws (smallest screws). Two from the top and two from bottom.
-- Slide over the first motor holder and fasten it using two M2x6mm screws (one on each side).
-- Install both motor horns, securing the top horn with a M3x6mm screw.
-- Attach the shoulder part.
-- Tighten the shoulder part with 4 M3x6mm screws on top and 4 M3x6mm screws on the bottom
-- Add the shoulder motor holder.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Joint1_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-### Joint 2
-
-- Slide the second motor in from the top.
-- Fasten the second motor with 4 M2x6mm screws.
-- Attach both motor horns to motor 2, again use the M3x6mm horn screw.
-- Attach the upper arm with 4 M3x6mm screws on each side.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Joint2_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-### Joint 3
-
-- Insert motor 3 and fasten using 4 M2x6mm screws
-- Attach both motor horns to motor 3 and secure one again with a M3x6mm horn screw.
-- Connect the forearm to motor 3 using 4 M3x6mm screws on each side.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Joint3_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-### Joint 4
-
-- Slide over motor holder 4.
-- Slide in motor 4.
-- Fasten motor 4 with 4 M2x6mm screws and attach its motor horns, use a M3x6mm horn screw.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Joint4_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-### Joint 5
-
-- Insert motor 5 into the wrist holder and secure it with 2 M2x6mm front screws.
-- Install only one motor horn on the wrist motor and secure it with a M3x6mm horn screw.
-- Secure the wrist to motor 4 using 4 M3x6mm screws on both sides.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Joint5_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-### Gripper / Handle
-
-<hfoptions id="assembly">
-<hfoption id="Follower">
-
-- Attach the gripper to motor 5, attach it to the motor horn on the wrist using 4 M3x6mm screws.
-- Insert the gripper motor and secure it with 2 M2x6mm screws on each side.
-- Attach the motor horns and again use a M3x6mm horn screw.
-- Install the gripper claw and secure it with 4 M3x6mm screws on both sides.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Gripper_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-</hfoption>
-<hfoption id="Leader">
-
-- Mount the leader holder onto the wrist and secure it with 4 M3x6mm screws.
-- Attach the handle to motor 5 using 1 M2x6mm screw.
-- Insert the gripper motor, secure it with 2 M2x6mm screws on each side, attach a motor horn using a M3x6mm horn screw.
-- Attach the follower trigger with 4 M3x6mm screws.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Leader_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-</hfoption>
-</hfoptions>
-
-##### Wiring
-
-- Attach the motor controller on the back.
-- Then insert all wires, use the wire guides everywhere to make sure the wires don't unplug themself and stay in place.
-
-<div class="video-container">
-  <video controls width="600">
-    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/Wiring_v2.mp4" type="video/mp4" />
-  </video>
-</div>
-
-## Calibrate
-
-Next, you'll need to calibrate your SO-101 robot to ensure that the leader and follower arms have the same position values when they are in the same physical position.
-The calibration process is very important because it allows a neural network trained on one SO-101 robot to work on another.
-
-#### Manual calibration of follower arm
-
-You will need to move the follower arm to these positions sequentially, note that the rotated position is on the right side of the robot and you have to open the gripper fully.
-
-| 1. Middle position | 2. Zero position                                                                                                                                       | 3. Rotated position                                                                                                                                             | 4. Rest position                                                                                                                                       |
-| ------------ |------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/follower_middle.webp?raw=true" alt="SO-101 leader arm middle position" title="SO-101 leader arm middle position" style="width:100%;"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/follower_zero.webp?raw=true" alt="SO-101 leader arm zero position" title="SO-101 leader arm zero position" style="width:100%;"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/follower_rotated.webp?raw=true" alt="SO-101 leader arm rotated position" title="SO-101 leader arm rotated position" style="width:100%;"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/follower_rest.webp?raw=true" alt="SO-101 leader arm rest position" title="SO-101 leader arm rest position" style="width:100%;"> |
-
-Make sure both arms are connected and run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_follower"]'
-```
-
-#### Manual calibration of leader arm
-You will also need to move the leader arm to these positions sequentially:
-
-| 1. Middle position | 2. Zero position                                                                                                                                       | 3. Rotated position                                                                                                                                             | 4. Rest position                                                                                                                                       |
-| ------------ |------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/leader_middle.webp?raw=true" alt="SO-101 leader arm middle position" title="SO-101 leader arm middle position" style="width:100%;"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/leader_zero.webp?raw=true" alt="SO-101 leader arm zero position" title="SO-101 leader arm zero position" style="width:100%;"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/leader_rotated.webp?raw=true" alt="SO-101 leader arm rotated position" title="SO-101 leader arm rotated position" style="width:100%;"> | <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/leader_rest.webp?raw=true" alt="SO-101 leader arm rest position" title="SO-101 leader arm rest position" style="width:100%;"> |
-
-Run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_leader"]'
-```
-
-Congrats 🎉, your robot is all set to learn a task on its own. Start training it by following this tutorial: [Getting started with real-world robots](./getting_started_real_world_robot)

docs/source/async.mdx

@@ -0,0 +1,272 @@
+# Asynchronous Inference
+
+With our [SmolVLA](https://huggingface.co/papers/2506.01844) we introduced a new way to run inference on real-world robots, **decoupling action prediction from action execution**.
+In this tutorial, we'll show how to use asynchronous inference (_async inference_) using a finetuned version of SmolVLA, and all the policies supported by LeRobot.
+**Try async inference with all the policies** supported by LeRobot!
+
+**What you'll learn:**
+1. Why asynchronous inference matters and how it compares to, more traditional, sequential inference.
+2. How to spin-up a `PolicyServer` and connect a `RobotClient` from the same machine, and even over the network.
+3. How to tune key parameters (`actions_per_chunk`, `chunk_size_threshold`) for your robot and policy.
+
+If you get stuck, hop into our [Discord community](https://discord.gg/s3KuuzsPFb)!
+
+
+In a nutshell: with *async inference*, your robot keeps acting while the policy server is already busy computing the next chunk of actions---eliminating "wait-for-inference" lags and unlocking smoother, more reactive behaviours.
+This is fundamentally different from synchronous inference (sync), where the robot stays idle while the policy computes the next chunk of actions.
+
+---
+## Getting started with async inference
+
+You can read more information on asynchronous inference in our [blogpost](https://huggingface.co/blog/async-robot-inference). This guide is designed to help you quickly set up and run asynchronous inference in your environment.
+
+First, install `lerobot` with the `async` tag, to install the extra dependencies required to run async inference.
+
+```shell
+pip install -e ".[async]"
+```
+
+Then, spin up a policy server (in one terminal, or in a separate machine) specifying the host address and port for the client to connect to.
+You can spin up a policy server running:
+
+```shell
+python src/lerobot/scripts/server/policy_server.py \
+    --host=127.0.0.1 \
+    --port=8080 \
+```
+
+This will start a policy server listening on `127.0.0.1:8080` (`localhost`, port 8080). At this stage, the policy server is empty, as all information related to which policy to run and with which parameters are specified during the first handshake with the client. Spin up a client with:
+
+```shell
+python src/lerobot/scripts/server/robot_client.py \
+    --server_address=127.0.0.1:8080 \ # SERVER: the host address and port of the policy server
+    --robot.type=so100_follower \ # ROBOT: your robot type
+    --robot.port=/dev/tty.usbmodem585A0076841 \ # ROBOT: your robot port
+    --robot.id=follower_so100 \ # ROBOT: your robot id, to load calibration file
+    --robot.cameras="{ laptop: {type: opencv, index_or_path: 0, width: 1920, height: 1080, fps: 30}, phone: {type: opencv, index_or_path: 0, width: 1920, height: 1080, fps: 30}}" \ # POLICY: the cameras used to acquire frames, with keys matching the keys expected by the policy
+    --task="dummy" \ # POLICY: The task to run the policy on (`Fold my t-shirt`). Not necessarily defined for all policies, such as `act`
+    --policy_type=your_policy_type \ # POLICY: the type of policy to run (smolvla, act, etc)
+    --pretrained_name_or_path=user/model \ # POLICY: the model name/path on server to the checkpoint to run (e.g., lerobot/smolvla_base)
+    --policy_device=mps \ # POLICY: the device to run the policy on, on the server
+    --actions_per_chunk=50 \ # POLICY: the number of actions to output at once
+    --chunk_size_threshold=0.5 \ # CLIENT: the threshold for the chunk size before sending a new observation to the server
+    --aggregate_fn_name=weighted_average \ # CLIENT: the function to aggregate actions on overlapping portions
+    --debug_visualize_queue_size=True # CLIENT: whether to visualize the queue size at runtime
+```
+In summary, you need to specify instructions for:
+- `SERVER`: the address and port of the policy server
+- `ROBOT`: the type of robot to connect to, the port to connect to, and the local `id` of the robot
+- `POLICY`: the type of policy to run, and the model name/path on server to the checkpoint to run. You also need to specify which device should the sever be using, and how many actions to output at once (capped at the policy max actions value).
+- `CLIENT`: the threshold for the chunk size before sending a new observation to the server, and the function to aggregate actions on overlapping portions. Optionally, you can also visualize the queue size at runtime, to help you tune the `CLIENT` parameters.
+
+Importantly,
+- `actions_per_chunk` and `chunk_size_threshold` are key parameters to tune for your setup.
+- `aggregate_fn_name` is the function to aggregate actions on overlapping portions. You can either add a new one to a registry of functions, or add your own in `robot_client.py` (see [here](NOTE:addlinktoLOC))
+- `debug_visualize_queue_size` is a useful tool to tune the `CLIENT` parameters.
+
+Done! You should see your robot moving around by now 😉
+---
+
+## Async vs. synchronous inference
+
+Synchronous inference relies on interleaving action chunk prediction and action execution. This inherently results in *idle frames*, frames where the robot awaits idle the policy's output: a new action chunk.
+In turn, inference is plagued by evident real-time lags, where the robot simply stops acting due to the lack of available actions.
+With robotics models increasing in size, this problem risks becoming only more severe.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/async-inference/sync.png" width="80%"></img>
+</p>
+<p align="center"><i>Synchronous inference</i> makes the robot idle while the policy is computing the next chunk of actions.</p>
+
+To overcome this, we design async inference, a paradigm where action planning and execution are decoupled, resulting in (1) higher adaptability and, most importantly, (2) no idle frames.
+Crucially, with async inference, the next action chunk is computed *before* the current one is exhausted, resulting in no idleness.
+Higher adaptability is ensured by aggregating the different action chunks on overlapping portions, obtaining an up-to-date plan and a tighter control loop.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/async-inference/async.png" width="80%"></img>
+</p>
+<p align="center"><i>Asynchronous inference</i> results in no idleness because the next chunk is computed before the current chunk is exhausted.</p>
+
+
+---
+
+## Start the Policy Server
+
+Policy servers are wrappers around a `PreTrainedPolicy` interfacing them with observations coming from a robot client.
+Policy servers are initialized as empty containers which are populated with the requested policy specified in the initial handshake between the robot client and the policy server.
+As such, spinning up a policy server is as easy as specifying the host address and port. If you're running the policy server on the same machine as the robot client, you can use `localhost` as the host address.
+
+<hfoptions id="start_policy_server">
+<hfoption id="Command">
+```bash
+python -m lerobot.scripts.server.policy_server \
+    --host="localhost" \
+    --port=8080
+```
+</hfoption>
+<hfoption id="API example">
+```python
+from lerobot.scripts.server.configs import PolicyServerConfig
+from lerobot.scripts.server.policy_server import serve
+
+config = PolicyServerConfig(
+    host="localhost",
+    port=8080,
+)
+serve(config)
+```
+</hfoption>
+</hfoptions>
+
+This listens on `localhost:8080` for an incoming connection from the associated`RobotClient`, which will communicate which policy to run during the first client-server handshake.
+
+---
+
+## Launch the Robot Client
+
+`RobotClient` is a wrapper around a `Robot` instance, which `RobotClient` connects to the (possibly remote) `PolicyServer`.
+The `RobotClient` streams observations to the `PolicyServer`, and receives action chunks obtained running inference on the server (which we assume to have better computational resources than the robot controller).
+
+<hfoptions id="start_robot_client">
+<hfoption id="Command">
+```bash
+python src/lerobot/scripts/server/robot_client.py \
+    --server_address=127.0.0.1:8080 \ # SERVER: the host address and port of the policy server
+    --robot.type=so100_follower \ # ROBOT: your robot type
+    --robot.port=/dev/tty.usbmodem585A0076841 \ # ROBOT: your robot port
+    --robot.id=follower_so100 \ # ROBOT: your robot id, to load calibration file
+    --robot.cameras="{ laptop: {type: opencv, index_or_path: 0, width: 1920, height: 1080, fps: 30}, phone: {type: opencv, index_or_path: 0, width: 1920, height: 1080, fps: 30}}" \ # POLICY: the cameras used to acquire frames, with keys matching the keys expected by the policy
+    --task="dummy" \ # POLICY: The task to run the policy on (`Fold my t-shirt`). Not necessarily defined for all policies, such as `act`
+    --policy_type=your_policy_type \ # POLICY: the type of policy to run (smolvla, act, etc)
+    --pretrained_name_or_path=user/model \ # POLICY: the model name/path on server to the checkpoint to run (e.g., lerobot/smolvla_base)
+    --policy_device=mps \ # POLICY: the device to run the policy on, on the server
+    --actions_per_chunk=50 \ # POLICY: the number of actions to output at once
+    --chunk_size_threshold=0.5 \ # CLIENT: the threshold for the chunk size before sending a new observation to the server
+    --aggregate_fn_name=weighted_average \ # CLIENT: the function to aggregate actions on overlapping portions
+    --debug_visualize_queue_size=True # CLIENT: whether to visualize the queue size at runtime
+```
+</hfoption>
+<hfoption id="API example">
+```python
+import threading
+from lerobot.robots.so100_follower import SO100FollowerConfig
+from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.scripts.server.configs import RobotClientConfig
+from lerobot.scripts.server.robot_client import RobotClient
+from lerobot.scripts.server.helpers import visualize_action_queue_size
+
+# 1. Create the robot instance
+"""Check out the cameras available in your setup by running `python lerobot/find_cameras.py`"""
+# these cameras must match the ones expected by the policy
+# check the config.json on the Hub for the policy you are using
+camera_cfg = {
+    "top": OpenCVCameraConfig(index_or_path=0, width=640, height=480, fps=30),
+    "side": OpenCVCameraConfig(index_or_path=1, width=640, height=480, fps=30)
+}
+
+robot_cfg = SO100FollowerConfig(
+  port="/dev/tty.usbmodem585A0076841",
+  id="follower_so100",
+  cameras=camera_cfg
+)
+
+# 3. Create client configuration
+client_cfg = RobotClientConfig(
+    robot=robot_cfg,
+    server_address="localhost:8080",
+    policy_device="mps",
+    policy_type="smolvla",
+    pretrained_name_or_path="fracapuano/smolvla_async",
+    chunk_size_threshold=0.5,
+    actions_per_chunk=50,  # make sure this is less than the max actions of the policy
+)
+
+# 4. Create and start client
+client = RobotClient(client_cfg)
+
+# 5. Specify the task
+task = "Don't do anything, stay still"
+
+if client.start():
+    # Start action receiver thread
+    action_receiver_thread = threading.Thread(target=client.receive_actions, daemon=True)
+    action_receiver_thread.start()
+
+    try:
+        # Run the control loop
+        client.control_loop(task)
+    except KeyboardInterrupt:
+        client.stop()
+        action_receiver_thread.join()
+        # (Optionally) plot the action queue size
+        visualize_action_queue_size(client.action_queue_size)
+```
+</hfoption>
+</hfoptions>
+
+The following two parameters are key in every setup:
+
+<table>
+  <thead>
+    <tr>
+      <th>Hyperparameter</th>
+      <th>Default</th>
+      <th>What it does</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td><code>actions_per_chunk</code></td>
+      <td>50</td>
+      <td>How many actions the policy outputs at once. Typical values: 10-50.</td>
+    </tr>
+    <tr>
+      <td><code>chunk_size_threshold</code></td>
+      <td>0.7</td>
+      <td>When the queue is ≤ 50% full, the client sends a fresh observation. Value in [0, 1].</td>
+    </tr>
+  </tbody>
+</table>
+
+<Tip>
+Different values of `actions_per_chunk` and `chunk_size_threshold` do result in different behaviours.
+</Tip>
+
+On the one hand, increasing the value of `actions_per_chunk` will result in reducing the likelihood of ending up with no actions to execute, as more actions will be available when the new chunk is computed.
+However, larger values of `actions_per_chunk` might also result in less precise actions, due to the compounding errors consequent to predicting actions over longer timespans.
+
+On the other hand, increasing the value of `chunk_size_threshold` will result in sending out to the `PolicyServer` observations for inference more often, resulting in a larger number of updates action chunks, overlapping on significant portions. This results in high adaptability, in the limit predicting one action chunk for each observation, which is in turn only marginally consumed while a new one is produced.
+This option does also put more pressure on the inference pipeline, as a consequence of the many requests. Conversely, values of `chunk_size_threshold` close to 0.0 collapse to the synchronous edge case, whereby new observations are only sent out whenever the current chunk is exhausted.
+
+We found the default values of `actions_per_chunk` and `chunk_size_threshold` to work well in the experiments we developed for the [SmolVLA paper](https://huggingface.co/papers/2506.01844), but recommend experimenting with different values to find the best fit for your setup.
+
+### Tuning async inference for your setup
+
+1. **Choose your computational resources carefully.** [PI0](https://huggingface.co/lerobot/pi0) occupies 14GB of memory at inference time, while [SmolVLA](https://huggingface.co/lerobot/smolvla_base) requires only ~2GB. You should identify the best computational resource for your use case keeping in mind smaller policies require less computational resources. The combination of policy and device used (CPU-intensive, using MPS, or the number of CUDA cores on a given NVIDIA GPU) directly impacts the average inference latency you should expect.
+2. **Adjust your `fps` based on inference latency.** While the server generates a new action chunk, the client is not idle and is stepping through its current action queue. If the two processes happen at fundamentally different speeds, the client might end up with an empty queue. As such, you should reduce your fps if you consistently run out of actions in queue.
+3. **Adjust `chunk_size_threshold`**.
+   - Values closer to `0.0` result in almost sequential behavior. Values closer to `1.0` → send observation every step (more bandwidth, relies on good world-model).
+   - We found values around 0.5-0.6 to work well. If you want to tweak this, spin up a `RobotClient` setting the `--debug-visualize-queue-size` to `True`. This will plot the action queue size evolution at runtime, and you can use it to find the value of `chunk_size_threshold` that works best for your setup.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/async-inference/queues.png" width="80%"></img>
+</p>
+<p align="center"><i>The action queue size is plotted at runtime when the `--debug-visualize-queue-size` flag is passed, for various levels of `chunk_size_threshold` (`g` in the SmolVLA paper).</i></p>
+
+
+---
+
+## Conclusion
+
+Asynchronous inference represents a significant advancement in real-time robotics control, addressing the fundamental challenge of inference latency that has long plagued robotics applications. Through this tutorial, you've learned how to implement a complete async inference pipeline that eliminates idle frames and enables smoother, more reactive robot behaviors.
+
+**Key Takeaways:**
+
+- **Paradigm Shift**: Async inference decouples action prediction from execution, allowing robots to continue acting while new action chunks are computed in parallel
+- **Performance Benefits**: Eliminates "wait-for-inference" lags that are inherent in synchronous approaches, becoming increasingly important as policy models grow larger
+- **Flexible Architecture**: The server-client design enables distributed computing, where inference can run on powerful remote hardware while maintaining real-time robot control
+- **Tunable Parameters**: Success depends on properly configuring `actions_per_chunk` and `chunk_size_threshold` for your specific hardware, policy, and task requirements
+- **Universal Compatibility**: Works with all LeRobot-supported policies, from lightweight ACT models to vision-language models like SmolVLA
+
+Start experimenting with the default parameters, monitor your action queue sizes, and iteratively refine your setup to achieve optimal performance for your specific use case.
+If you want to discuss this further, hop into our [Discord community](https://discord.gg/s3KuuzsPFb), or open an issue on our [GitHub repository](https://github.com/lerobot/lerobot/issues).

docs/source/backwardcomp.mdx

@@ -0,0 +1,82 @@
+# Backward compatibility
+
+## Hardware API redesign
+
+PR [#777](https://github.com/huggingface/lerobot/pull/777) improves the LeRobot calibration but is **not backward-compatible**. Below is a overview of what changed and how you can continue to work with datasets created before this pull request.
+
+### What changed?
+
+|                                   | Before PR #777                                    | After PR #777                                                               |
+| --------------------------------- | ------------------------------------------------- | --------------------------------------------------------------------------- |
+| **Joint range**                   | Degrees `-180...180°`                              | **Normalised range** Joints: `–100...100` Gripper: `0...100` |
+| **Zero position (SO100 / SO101)** | Arm fully extended horizontally                   | **In middle of the range for each joint**                                        |
+| **Boundary handling**             | Software safeguards to detect ±180 ° wrap-arounds | No wrap-around logic needed due to mid-range zero                           |
+
+---
+
+### Impact on existing datasets
+
+* Recorded trajectories created **before** PR #777 will replay incorrectly if loaded directly:
+  * Joint angles are offset and incorrectly normalized.
+* Any models directly finetuned or trained on the old data will need their inputs and outputs converted.
+
+### Using datasets made with the previous calibration system
+We provide a migration example script for replaying an episode recorded with the previous calibration here: `examples/backward_compatibility/replay.py`.
+Below we take you through the modifications that are done in the example script to make the previous calibration datasets work.
+
+```diff
++   key = f"{name.removeprefix('main_')}.pos"
+    action[key] = action_array[i].item()
++   action["shoulder_lift.pos"] = -(action["shoulder_lift.pos"] - 90)
++   action["elbow_flex.pos"] -= 90
+```
+
+Let's break this down.
+New codebase uses `.pos` suffix for the position observations and we have removed `main_` prefix:
+```python
+key = f"{name.removeprefix('main_')}.pos"
+```
+
+For `"shoulder_lift"` (id = 2), the 0 position is changed by -90 degrees and the direction is reversed compared to old calibration/code.
+```python
+action["shoulder_lift.pos"] = -(action["shoulder_lift.pos"] - 90)
+```
+For `"elbow_flex"` (id = 3), the 0 position is changed by -90 degrees compared to old calibration/code.
+```python
+action["elbow_flex.pos"] -= 90
+```
+
+To use degrees normalization we then set the `--robot.use_degrees` option to `true`.
+```diff
+python examples/backward_compatibility/replay.py \
+    --robot.type=so101_follower \
+    --robot.port=/dev/tty.usbmodem5A460814411 \
+    --robot.id=blue \
++   --robot.use_degrees=true \
+    --dataset.repo_id=my_dataset_id \
+    --dataset.episode=0
+```
+
+### Using policies trained with the previous calibration system
+
+Policies output actions in the same format as the datasets (`torch.Tensors`). Therefore, the same transformations should be applied.
+
+To find these transformations, we recommend to first try and and replay an episode of the dataset your policy was trained on using the section above.
+Then, add these same transformations on your inference script (shown here in the `record.py` script):
+```diff
+action_values = predict_action(
+    observation_frame,
+    policy,
+    get_safe_torch_device(policy.config.device),
+    policy.config.use_amp,
+    task=single_task,
+    robot_type=robot.robot_type,
+    )
+    action = {key: action_values[i].item() for i, key in enumerate(robot.action_features)}
+
++   action["shoulder_lift.pos"] = -(action["shoulder_lift.pos"] - 90)
++   action["elbow_flex.pos"] -= 90
+    robot.send_action(action)
+```
+
+If you have questions or run into migration issues, feel free to ask them on [Discord](https://discord.gg/s3KuuzsPFb)

docs/source/cameras.mdx

@@ -0,0 +1,173 @@
+# Cameras
+
+LeRobot offers multiple options for video capture, including phone cameras, built-in laptop cameras, external webcams, and Intel RealSense cameras. To efficiently record frames from most cameras, you can use either the `OpenCVCamera` or `RealSenseCamera` class. For additional compatibility details on the `OpenCVCamera` class, refer to the [Video I/O with OpenCV Overview](https://docs.opencv.org/4.x/d0/da7/videoio_overview.html).
+
+### Finding your camera
+
+To instantiate a camera, you need a camera identifier. This identifier might change if you reboot your computer or re-plug your camera, a behavior mostly dependant on your operating system.
+
+To find the camera indices of the cameras plugged into your system, run the following script:
+```bash
+python -m lerobot.find_cameras opencv # or realsense for Intel Realsense cameras
+```
+
+The output will look something like this if you have two cameras connected:
+```
+--- Detected Cameras ---
+Camera #0:
+  Name: OpenCV Camera @ 0
+  Type: OpenCV
+  Id: 0
+  Backend api: AVFOUNDATION
+  Default stream profile:
+    Format: 16.0
+    Width: 1920
+    Height: 1080
+    Fps: 15.0
+--------------------
+(more cameras ...)
+```
+
+> [!WARNING]
+> When using Intel RealSense cameras in `macOS`, you could get this [error](https://github.com/IntelRealSense/librealsense/issues/12307): `Error finding RealSense cameras: failed to set power state`, this can be solved by running the same command with `sudo` permissions. Note that using RealSense cameras in `macOS` is unstable.
+
+
+## Use Cameras
+
+Below are two examples, demonstrating how to work with the API.
+
+- **Asynchronous frame capture** using an OpenCV-based camera
+- **Color and depth capture** using an Intel RealSense camera
+
+
+<hfoptions id="shell_restart">
+<hfoption id="Open CV Camera">
+
+```python
+from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.cameras.opencv.camera_opencv import OpenCVCamera
+from lerobot.cameras.configs import ColorMode, Cv2Rotation
+
+# Construct an `OpenCVCameraConfig` with your desired FPS, resolution, color mode, and rotation.
+config = OpenCVCameraConfig(
+    index_or_path=0,
+    fps=15,
+    width=1920,
+    height=1080,
+    color_mode=ColorMode.RGB,
+    rotation=Cv2Rotation.NO_ROTATION
+)
+
+# Instantiate and connect an `OpenCVCamera`, performing a warm-up read (default).
+camera = OpenCVCamera(config)
+camera.connect()
+
+# Read frames asynchronously in a loop via `async_read(timeout_ms)`
+try:
+    for i in range(10):
+        frame = camera.async_read(timeout_ms=200)
+        print(f"Async frame {i} shape:", frame.shape)
+finally:
+    camera.disconnect()
+```
+
+</hfoption>
+<hfoption id="Intel Realsense Camera">
+
+```python
+from lerobot.cameras.realsense.configuration_realsense import RealSenseCameraConfig
+from lerobot.cameras.realsense.camera_realsense import RealSenseCamera
+from lerobot.cameras.configs import ColorMode, Cv2Rotation
+
+# Create a `RealSenseCameraConfig` specifying your camera’s serial number and enabling depth.
+config = RealSenseCameraConfig(
+    serial_number_or_name="233522074606",
+    fps=15,
+    width=640,
+    height=480,
+    color_mode=ColorMode.RGB,
+    use_depth=True,
+    rotation=Cv2Rotation.NO_ROTATION
+)
+
+# Instantiate and connect a `RealSenseCamera` with warm-up read (default).
+camera = RealSenseCamera(config)
+camera.connect()
+
+# Capture a color frame via `read()` and a depth map via `read_depth()`.
+try:
+    color_frame = camera.read()
+    depth_map = camera.read_depth()
+    print("Color frame shape:", color_frame.shape)
+    print("Depth map shape:", depth_map.shape)
+finally:
+    camera.disconnect()
+```
+</hfoption>
+</hfoptions>
+
+
+## Use your phone
+<hfoptions id="use phone">
+<hfoption id="Mac">
+
+To use your iPhone as a camera on macOS, enable the Continuity Camera feature:
+- Ensure your Mac is running macOS 13 or later, and your iPhone is on iOS 16 or later.
+- Sign in both devices with the same Apple ID.
+- Connect your devices with a USB cable or turn on Wi-Fi and Bluetooth for a wireless connection.
+
+For more details, visit [Apple support](https://support.apple.com/en-gb/guide/mac-help/mchl77879b8a/mac).
+
+Your iPhone should be detected automatically when running the camera setup script in the next section.
+
+</hfoption>
+<hfoption id="Linux">
+
+If you want to use your phone as a camera on Linux, follow these steps to set up a virtual camera
+
+1. *Install `v4l2loopback-dkms` and `v4l-utils`*. Those packages are required to create virtual camera devices (`v4l2loopback`) and verify their settings with the `v4l2-ctl` utility from `v4l-utils`. Install them using:
+```python
+sudo apt install v4l2loopback-dkms v4l-utils
+```
+2. *Install [DroidCam](https://droidcam.app) on your phone*. This app is available for both iOS and Android.
+3. *Install [OBS Studio](https://obsproject.com)*. This software will help you manage the camera feed. Install it using [Flatpak](https://flatpak.org):
+```python
+flatpak install flathub com.obsproject.Studio
+```
+4. *Install the DroidCam OBS plugin*. This plugin integrates DroidCam with OBS Studio. Install it with:
+```python
+flatpak install flathub com.obsproject.Studio.Plugin.DroidCam
+```
+5. *Start OBS Studio*. Launch with:
+```python
+flatpak run com.obsproject.Studio
+```
+6. *Add your phone as a source*. Follow the instructions [here](https://droidcam.app/obs/usage). Be sure to set the resolution to `640x480`.
+7. *Adjust resolution settings*. In OBS Studio, go to `File > Settings > Video`. Change the `Base(Canvas) Resolution` and the `Output(Scaled) Resolution` to `640x480` by manually typing it in.
+8. *Start virtual camera*. In OBS Studio, follow the instructions [here](https://obsproject.com/kb/virtual-camera-guide).
+9. *Verify the virtual camera setup*. Use `v4l2-ctl` to list the devices:
+```python
+v4l2-ctl --list-devices
+```
+You should see an entry like:
+```
+VirtualCam (platform:v4l2loopback-000):
+/dev/video1
+```
+10. *Check the camera resolution*. Use `v4l2-ctl` to ensure that the virtual camera output resolution is `640x480`. Change `/dev/video1` to the port of your virtual camera from the output of `v4l2-ctl --list-devices`.
+```python
+v4l2-ctl -d /dev/video1 --get-fmt-video
+```
+You should see an entry like:
+```
+>>> Format Video Capture:
+>>>	Width/Height      : 640/480
+>>>	Pixel Format      : 'YUYV' (YUYV 4:2:2)
+```
+
+Troubleshooting: If the resolution is not correct you will have to delete the Virtual Camera port and try again as it cannot be changed.
+
+If everything is set up correctly, you can proceed with the rest of the tutorial.
+
+</hfoption>
+</hfoptions>

docs/source/contributing.md

@@ -0,0 +1 @@
+../../CONTRIBUTING.md

docs/source/getting_started_real_world_robot.mdx

@@ -1,370 +0,0 @@
-# Getting Started with Real-World Robots
-
-This tutorial will explain you how to train a neural network to autonomously control a real robot.
-
-**You'll learn:**
-1. How to record and visualize your dataset.
-2. How to train a policy using your data and prepare it for evaluation.
-3. How to evaluate your policy and visualize the results.
-
-By following these steps, you'll be able to replicate tasks like picking up a Lego block and placing it in a bin with a high success rate, as demonstrated in [this video](https://x.com/RemiCadene/status/1814680760592572934).
-
-This tutorial is specifically made for the affordable [SO-101](https://github.com/TheRobotStudio/SO-ARM100) robot, but it contains additional information to be easily adapted to various types of robots like [Aloha bimanual robot](https://aloha-2.github.io) by changing some configurations. The SO-101 consists of a leader arm and a follower arm, each with 6 motors. It can work with one or several cameras to record the scene, which serve as visual sensors for the robot.
-
-During the data collection phase, you will control the follower arm by moving the leader arm. This process is known as "teleoperation." This technique is used to collect robot trajectories. Afterward, you'll train a neural network to imitate these trajectories and deploy the network to enable your robot to operate autonomously.
-
-If you encounter any issues at any step of the tutorial, feel free to seek help on [Discord](https://discord.com/invite/s3KuuzsPFb) or don't hesitate to iterate with us on the tutorial by creating issues or pull requests.
-
-## Setup and Calibrate
-
-If you haven't yet setup and calibrate the SO-101 follow these steps:
-1. [Find ports and update config file](./assemble_so101#find-the-usb-ports-associated-to-each-arm)
-2. [Calibrate](./assemble_so101#calibrate)
-
-## Teleoperate
-
-Run this simple script to teleoperate your robot (it won't connect and display the cameras):
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --robot.cameras='{}' \
-  --control.type=teleoperate
-```
-
-The teleoperate command will automatically:
-1. Identify any missing calibrations and initiate the calibration procedure.
-2. Connect the robot and start teleoperation.
-
-## Setup Cameras
-
-To connect a camera you have three options:
-1. OpenCVCamera which allows us to use any camera: usb, realsense, laptop webcam
-2. iPhone camera with MacOS
-3. Phone camera on Linux
-
-### Use OpenCVCamera
-
-The [`OpenCVCamera`](../lerobot/common/robot_devices/cameras/opencv.py) class allows you to efficiently record frames from most cameras using the [`opencv2`](https://docs.opencv.org) library.  For more details on compatibility, see [Video I/O with OpenCV Overview](https://docs.opencv.org/4.x/d0/da7/videoio_overview.html).
-
-To instantiate an [`OpenCVCamera`](../lerobot/common/robot_devices/cameras/opencv.py), you need a camera index (e.g. `OpenCVCamera(camera_index=0)`). When you only have one camera like a webcam of a laptop, the camera index is usually `0` but it might differ, and the camera index might change if you reboot your computer or re-plug your camera. This behavior depends on your operating system.
-
-To find the camera indices, run the following utility script, which will save a few frames from each detected camera:
-```bash
-python lerobot/common/robot_devices/cameras/opencv.py \
-    --images-dir outputs/images_from_opencv_cameras
-```
-
-The output will look something like this if you have two cameras connected:
-```
-Mac or Windows detected. Finding available camera indices through scanning all indices from 0 to 60
-[...]
-Camera found at index 0
-Camera found at index 1
-[...]
-Connecting cameras
-OpenCVCamera(0, fps=30.0, width=1920.0, height=1080.0, color_mode=rgb)
-OpenCVCamera(1, fps=24.0, width=1920.0, height=1080.0, color_mode=rgb)
-Saving images to outputs/images_from_opencv_cameras
-Frame: 0000	Latency (ms): 39.52
-[...]
-Frame: 0046	Latency (ms): 40.07
-Images have been saved to outputs/images_from_opencv_cameras
-```
-
-Check the saved images in `outputs/images_from_opencv_cameras` to identify which camera index corresponds to which physical camera (e.g. `0` for `camera_00` or `1` for `camera_01`):
-```
-camera_00_frame_000000.png
-[...]
-camera_00_frame_000047.png
-camera_01_frame_000000.png
-[...]
-camera_01_frame_000047.png
-```
-
-Note: Some cameras may take a few seconds to warm up, and the first frame might be black or green.
-
-Now that you have the camera indexes, you should specify the camera's in the config. TODO(pepijn): add more info about setting camera config, rotate etc..
-
-### Use your phone
-<hfoptions id="use phone">
-<hfoption id="Mac">
-
-To use your iPhone as a camera on macOS, enable the Continuity Camera feature:
-- Ensure your Mac is running macOS 13 or later, and your iPhone is on iOS 16 or later.
-- Sign in both devices with the same Apple ID.
-- Connect your devices with a USB cable or turn on Wi-Fi and Bluetooth for a wireless connection.
-
-For more details, visit [Apple support](https://support.apple.com/en-gb/guide/mac-help/mchl77879b8a/mac).
-
-Your iPhone should be detected automatically when running the camera setup script in the next section.
-
-</hfoption>
-<hfoption id="Linux">
-
-If you want to use your phone as a camera on Linux, follow these steps to set up a virtual camera
-
-1. *Install `v4l2loopback-dkms` and `v4l-utils`*. Those packages are required to create virtual camera devices (`v4l2loopback`) and verify their settings with the `v4l2-ctl` utility from `v4l-utils`. Install them using:
-```python
-sudo apt install v4l2loopback-dkms v4l-utils
-```
-2. *Install [DroidCam](https://droidcam.app) on your phone*. This app is available for both iOS and Android.
-3. *Install [OBS Studio](https://obsproject.com)*. This software will help you manage the camera feed. Install it using [Flatpak](https://flatpak.org):
-```python
-flatpak install flathub com.obsproject.Studio
-```
-4. *Install the DroidCam OBS plugin*. This plugin integrates DroidCam with OBS Studio. Install it with:
-```python
-flatpak install flathub com.obsproject.Studio.Plugin.DroidCam
-```
-5. *Start OBS Studio*. Launch with:
-```python
-flatpak run com.obsproject.Studio
-```
-6. *Add your phone as a source*. Follow the instructions [here](https://droidcam.app/obs/usage). Be sure to set the resolution to `640x480`.
-7. *Adjust resolution settings*. In OBS Studio, go to `File > Settings > Video`. Change the `Base(Canvas) Resolution` and the `Output(Scaled) Resolution` to `640x480` by manually typing it in.
-8. *Start virtual camera*. In OBS Studio, follow the instructions [here](https://obsproject.com/kb/virtual-camera-guide).
-9. *Verify the virtual camera setup*. Use `v4l2-ctl` to list the devices:
-```python
-v4l2-ctl --list-devices
-```
-You should see an entry like:
-```
-VirtualCam (platform:v4l2loopback-000):
-/dev/video1
-```
-10. *Check the camera resolution*. Use `v4l2-ctl` to ensure that the virtual camera output resolution is `640x480`. Change `/dev/video1` to the port of your virtual camera from the output of `v4l2-ctl --list-devices`.
-```python
-v4l2-ctl -d /dev/video1 --get-fmt-video
-```
-You should see an entry like:
-```
->>> Format Video Capture:
->>>	Width/Height      : 640/480
->>>	Pixel Format      : 'YUYV' (YUYV 4:2:2)
-```
-
-Troubleshooting: If the resolution is not correct you will have to delete the Virtual Camera port and try again as it cannot be changed.
-
-If everything is set up correctly, you can proceed with the rest of the tutorial.
-
-</hfoption>
-</hfoptions>
-
-## Teleoperate with cameras
-
-We can now teleoperate again while at the same time visualizing the camera's and joint positions with `rerun`.
-
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=teleoperate
-  --control.display_data=true
-```
-
-## Record a dataset
-
-Once you're familiar with teleoperation, you can record your first dataset with SO-101.
-
-We use the Hugging Face hub features for uploading your dataset. If you haven't previously used the Hub, make sure you can login via the cli using a write-access token, this token can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens).
-
-Add your token to the cli by running this command:
-```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
-```
-
-Then store your Hugging Face repository name in a variable:
-```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
-echo $HF_USER
-```
-
-Now you can record a dataset, to record 2 episodes and upload your dataset to the hub execute this command:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/so101_test \
-  --control.tags='["so101","tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=2 \
-  --control.push_to_hub=true
-```
-
-You will see a lot of lines appearing like this one:
-```
-INFO 2024-08-10 15:02:58 ol_robot.py:219 dt:33.34 (30.0hz) dtRlead: 5.06 (197.5hz) dtWfoll: 0.25 (3963.7hz) dtRfoll: 6.22 (160.7hz) dtRlaptop: 32.57 (30.7hz) dtRphone: 33.84 (29.5hz)
-```
-
-| Field | Meaning |
-|:---|:---|
-| `2024-08-10 15:02:58` | Timestamp when `print` was called. |
-| `ol_robot.py:219` | Source file and line number of the `print` call (`lerobot/scripts/control_robot.py` at line `219`). |
-| `dt: 33.34 (30.0 Hz)` | Delta time (ms) between teleop steps (target: 30.0 Hz, `--fps 30`). Yellow if step is too slow. |
-| `dtRlead: 5.06 (197.5 Hz)` | Delta time (ms) for reading present position from the **leader arm**. |
-| `dtWfoll: 0.25 (3963.7 Hz)` | Delta time (ms) for writing goal position to the **follower arm** (asynchronous). |
-| `dtRfoll: 6.22 (160.7 Hz)` | Delta time (ms) for reading present position from the **follower arm**. |
-| `dtRlaptop: 32.57 (30.7 Hz)` | Delta time (ms) for capturing an image from the **laptop camera** (async thread). |
-| `dtRphone: 33.84 (29.5 Hz)` | Delta time (ms) for capturing an image from the **phone camera** (async thread). |
-
-
-#### Dataset upload
-Locally your dataset is stored in this folder: `~/.cache/huggingface/lerobot/{repo-id}` (e.g. `data/cadene/so101_test`). At the end of data recording, your dataset will be uploaded on your Hugging Face page (e.g. https://huggingface.co/datasets/cadene/so101_test) that you can obtain by running:
-```bash
-echo https://huggingface.co/datasets/${HF_USER}/so101_test
-```
-Your dataset will be automatically tagged with `LeRobot` for the community to find it easily, and you can also add custom tags (in this case `tutorial` for example).
-
-You can look for other LeRobot datasets on the hub by searching for `LeRobot` [tags](https://huggingface.co/datasets?other=LeRobot).
-
-#### Record function
-
-The `record` function provides a suite of tools for capturing and managing data during robot operation:
-
-##### 1. Frame Capture and Video Encoding
-- Frames from cameras are saved to disk during recording.
-- At the end of each episode, frames are encoded into video files.
-
-##### 2. Data Storage
-- Data is stored using the `LeRobotDataset` format.
-- By default, the dataset is pushed to your Hugging Face page.
-  - To disable uploading, use `--control.push_to_hub=false`.
-
-##### 3. Checkpointing and Resuming
-- Checkpoints are automatically created during recording.
-- If an issue occurs, you can resume by re-running the same command with `--control.resume=true`.
-- To start recording from scratch, **manually delete** the dataset directory.
-
-##### 4. Recording Parameters
-Set the flow of data recording using command-line arguments:
-- `--control.warmup_time_s=10`
-  Number of seconds before starting data collection (default: **10 seconds**).
-  Allows devices to warm up and synchronize.
-- `--control.episode_time_s=60`
-  Duration of each data recording episode (default: **60 seconds**).
-- `--control.reset_time_s=60`
-  Duration for resetting the environment after each episode (default: **60 seconds**).
-- `--control.num_episodes=50`
-  Total number of episodes to record (default: **50**).
-
-##### 5. Keyboard Controls During Recording
-Control the data recording flow using keyboard shortcuts:
-- Press **Right Arrow (`→`)**: Early stop the current episode or reset time and move to the next.
-- Press **Left Arrow (`←`)**: Cancel the current episode and re-record it.
-- Press **Escape (`ESC`)**: Immediately stop the session, encode videos, and upload the dataset.
-
-#### Tips for gathering data
-
-Once you're comfortable with data recording, you can create a larger dataset for training. A good starting task is grasping an object at different locations and placing it in a bin. We suggest recording at least 50 episodes, with 10 episodes per location. Keep the cameras fixed and maintain consistent grasping behavior throughout the recordings. Also make sure the object you are manipulating is visible on the camera's. A good rule of thumb is you should be able to do the task yourself by only looking at the camera images.
-
-In the following sections, you’ll train your neural network. After achieving reliable grasping performance, you can start introducing more variations during data collection, such as additional grasp locations, different grasping techniques, and altering camera positions.
-
-Avoid adding too much variation too quickly, as it may hinder your results.
-
-
-#### Troubleshooting:
-- On Linux, if the left and right arrow keys and escape key don't have any effect during data recording, make sure you've set the `$DISPLAY` environment variable. See [pynput limitations](https://pynput.readthedocs.io/en/latest/limitations.html#linux).
-
-## Visualize a dataset
-
-If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:
-```bash
-echo ${HF_USER}/so101_test
-```
-
-If you didn't upload with `--control.push_to_hub=false`, you can visualize it locally with (via a window in the browser `http://127.0.0.1:9090` with the visualization tool):
-```bash
-python lerobot/scripts/visualize_dataset_html.py \
-  --repo-id ${HF_USER}/so101_test \
-  --local-files-only 1
-```
-
-This will launch a local web server that looks like this:
-<div style="text-align:center;">
-  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/visualize_dataset_html.webp?raw=true" alt="Koch v1.1 leader and follower arms" title="Koch v1.1 leader and follower arms" width="100%"></img>
-</div>
-
-## Replay an episode
-
-A useful feature is the `replay` function, which allows to replay on your robot any episode that you've recorded or episodes from any dataset out there. This function helps you test the repeatability of your robot's actions and assess transferability across robots of the same model.
-
-You can replay the first episode on your robot with:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=replay \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/so101_test \
-  --control.episode=0
-```
-
-Your robot should replicate movements similar to those you recorded. For example, check out [this video](https://x.com/RemiCadene/status/1793654950905680090) where we use `replay` on a Aloha robot from [Trossen Robotics](https://www.trossenrobotics.com).
-
-## Train a policy
-
-To train a policy to control your robot, use the [`python lerobot/scripts/train.py`](../lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/so101_test \
-  --policy.type=act \
-  --output_dir=outputs/train/act_so101_test \
-  --job_name=act_so101_test \
-  --policy.device=cuda \
-  --wandb.enable=true
-```
-
-Let's explain the command:
-1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/so101_test`.
-2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../lerobot/common/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
-4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
-5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
-
-Training should take several hours. You will find checkpoints in `outputs/train/act_so101_test/checkpoints`.
-
-To resume training from a checkpoint, below is an example command to resume from `last` checkpoint of the `act_so101_test` policy:
-```bash
-python lerobot/scripts/train.py \
-  --config_path=outputs/train/act_so101_test/checkpoints/last/pretrained_model/train_config.json \
-  --resume=true
-```
-
-#### Upload policy checkpoints
-
-Once training is done, upload the latest checkpoint with:
-```bash
-huggingface-cli upload ${HF_USER}/act_so101_test \
-  outputs/train/act_so101_test/checkpoints/last/pretrained_model
-```
-
-You can also upload intermediate checkpoints with:
-```bash
-CKPT=010000
-huggingface-cli upload ${HF_USER}/act_so101_test${CKPT} \
-  outputs/train/act_so101_test/checkpoints/${CKPT}/pretrained_model
-```
-
-## Evaluate your policy
-
-You can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) but with a policy checkpoint as input. For instance, run this command to record 10 evaluation episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/eval_act_so101_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=10 \
-  --control.push_to_hub=true \
-  --control.policy.path=outputs/train/act_so101_test/checkpoints/last/pretrained_model
-```
-
-As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
-1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_act_so101_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_so101_test`).
-2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_so101_test`).

docs/source/hilserl.mdx

@@ -0,0 +1,548 @@
+# HIL-SERL Real Robot Training Workflow Guide
+
+In this tutorial you will go through the full Human-in-the-Loop Sample-Efficient Reinforcement Learning (HIL-SERL) workflow using LeRobot. You will master training a policy with RL on a real robot in just a few hours.
+
+HIL-SERL is a sample-efficient reinforcement learning algorithm that combines human demonstrations with online learning and human interventions. The approach starts from a small set of human demonstrations, uses them to train a reward classifier, and then employs an actor-learner architecture where humans can intervene during policy execution to guide exploration and correct unsafe behaviors. In this tutorial, you'll use a gamepad to provide interventions and control the robot during the learning process.
+
+It combines three key ingredients:
+	1.	**Offline demonstrations & reward classifier:** a handful of human-teleop episodes plus a vision-based success detector give the policy a shaped starting point.
+	2.	**On-robot actor / learner loop with human interventions:** a distributed Soft Actor Critic (SAC) learner updates the policy while an actor explores on the physical robot; the human can jump in at any time to correct dangerous or unproductive behaviour.
+	3.	**Safety & efficiency tools:** joint/end-effector (EE) bounds, crop region of interest (ROI) preprocessing and WandB monitoring keep the data useful and the hardware safe.
+
+Together these elements let HIL-SERL reach near-perfect task success and faster cycle times than imitation-only baselines.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/hilserl-main-figure.png" alt="HIL-SERL workflow" title="HIL-SERL workflow" width="100%"></img>
+</p>
+
+<p align="center"><i>HIL-SERL workflow, Luo et al. 2024</i></p>
+
+This guide provides step-by-step instructions for training a robot policy using LeRobot's HilSerl implementation to train on a real robot.
+
+## What do I need?
+
+- A gamepad (recommended) or keyboard to control the robot
+- A Nvidia GPU
+- A real robot with a follower and leader arm (optional if you use the keyboard or the gamepad)
+- A URDF file for the robot for the kinematics package (check `lerobot/common/model/kinematics.py`)
+
+## What kind of tasks can I train?
+
+One can use HIL-SERL to train on a variety of manipulation tasks. Some recommendations:
+- Start with a simple task to understand how the system works.
+  - Push cube to a goal region
+  - Pick and lift cube with the gripper
+- Avoid extremely long horizon tasks. Focus on tasks that can be completed in 5-10 seconds.
+- Once you have a good idea of how the system works, you can try more complex tasks and longer horizons.
+  - Pick and place cube
+  - Bimanual tasks to pick objects with two arms
+  - Hand-over tasks to transfer objects from one arm to another
+  - Go crazy!
+
+## Install LeRobot with HIL-SERL
+
+To install LeRobot with HIL-SERL, you need to install the `hilserl` extra.
+
+```bash
+pip install -e ".[hilserl]"
+```
+
+## Real Robot Training Workflow
+
+### Understanding Configuration
+
+The training process begins with proper configuration for the HILSerl environment. The configuration class of interest is `HILSerlRobotEnvConfig` in `lerobot/envs/configs.py`. Which is defined as:
+
+```python
+class HILSerlRobotEnvConfig(EnvConfig):
+    robot: RobotConfig | None = None    # Main robot agent (defined in `lerobot/robots`)
+    teleop: TeleoperatorConfig | None = None    # Teleoperator agent, e.g., gamepad or leader arm, (defined in `lerobot/teleoperators`)
+    wrapper: EnvTransformConfig | None = None    # Environment wrapper settings; check `lerobot/scripts/server/gym_manipulator.py`
+    fps: int = 10    # Control frequency
+    name: str = "real_robot"    # Environment name
+    mode: str = None    # "record", "replay", or None (for training)
+    repo_id: str | None = None    # LeRobot dataset repository ID
+    dataset_root: str | None = None    # Local dataset root (optional)
+    task: str = ""    # Task identifier
+    num_episodes: int = 10    # Number of episodes for recording
+    episode: int = 0    # episode index for replay
+    device: str = "cuda"    # Compute device
+    push_to_hub: bool = True    # Whether to push the recorded datasets to Hub
+    pretrained_policy_name_or_path: str | None = None    # For policy loading
+    reward_classifier_pretrained_path: str | None = None    # For reward model
+    number_of_steps_after_success: int = 0    # For reward classifier, collect more positive examples after a success to train a classifier
+```
+
+
+### Finding Robot Workspace Bounds
+
+Before collecting demonstrations, you need to determine the appropriate operational bounds for your robot.
+
+This helps simplify the problem of learning on the real robot in two ways: 1) by limiting the robot's operational space to a specific region that solves the task and avoids unnecessary or unsafe exploration, and 2) by allowing training in end-effector space rather than joint space. Empirically, learning in joint space for reinforcement learning in manipulation is often a harder problem - some tasks are nearly impossible to learn in joint space but become learnable when the action space is transformed to end-effector coordinates.
+
+**Using find_joint_limits.py**
+
+This script helps you find the safe operational bounds for your robot's end-effector. Given that you have a follower and leader arm, you can use the script to find the bounds for the follower arm that will be applied during training.
+Bounding the action space will reduce the redundant exploration of the agent and guarantees safety.
+
+```bash
+python -m lerobot.scripts.find_joint_limits \
+    --robot.type=so100_follower \
+    --robot.port=/dev/tty.usbmodem58760431541 \
+    --robot.id=black \
+    --teleop.type=so100_leader \
+    --teleop.port=/dev/tty.usbmodem58760431551 \
+    --teleop.id=blue
+```
+
+**Workflow**
+
+1. Run the script and move the robot through the space that solves the task
+2. The script will record the minimum and maximum end-effector positions and the joint angles and prints them to the console, for example:
+   ```
+   Max ee position [0.2417 0.2012 0.1027]
+   Min ee position [0.1663 -0.0823 0.0336]
+   Max joint positions [-20.0, -20.0, -20.0, -20.0, -20.0, -20.0]
+   Min joint positions [50.0, 50.0, 50.0, 50.0, 50.0, 50.0]
+   ```
+3. Use these values in the configuration of your teleoperation device (TeleoperatorConfig) under the `end_effector_bounds` field
+
+**Example Configuration**
+
+```json
+"end_effector_bounds": {
+    "max": [0.24, 0.20, 0.10],
+    "min": [0.16, -0.08, 0.03]
+}
+```
+
+### Collecting Demonstrations
+
+With the bounds defined, you can safely collect demonstrations for training. Training RL with off-policy algorithm allows us to use offline datasets collected in order to improve the efficiency of the learning process.
+
+**Setting Up Record Mode**
+
+Create a configuration file for recording demonstrations (or edit an existing one like [env_config_so100.json](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/env_config_so100.json)):
+
+1. Set `mode` to `"record"`
+2. Specify a unique `repo_id` for your dataset (e.g., "username/task_name")
+3. Set `num_episodes` to the number of demonstrations you want to collect
+4. Set `crop_params_dict` to `null` initially (we'll determine crops later)
+5. Configure `robot`, `cameras`, and other hardware settings
+
+Example configuration section:
+```json
+"mode": "record",
+"repo_id": "username/pick_lift_cube",
+"dataset_root": null,
+"task": "pick_and_lift",
+"num_episodes": 15,
+"episode": 0,
+"push_to_hub": true
+```
+
+### Using a Teleoperation Device
+
+Along with your robot, you will need a teleoperation device to control it in order to collect datasets of your task and perform interventions during the online training.
+We support using a gamepad or a keyboard or the leader arm of the robot.
+
+HIL-Serl learns actions in the end-effector space of the robot. Therefore, the teleoperation will control the end-effector's x,y,z displacements.
+
+For that we need to define a version of the robot that takes actions in the end-effector space. Check the robot class `SO100FollowerEndEffector` and its configuration `SO100FollowerEndEffectorConfig` for the default parameters related to the end-effector space.
+
+```python
+class SO100FollowerEndEffectorConfig(SO100FollowerConfig):
+    """Configuration for the SO100FollowerEndEffector robot."""
+
+    # Default bounds for the end-effector position (in meters)
+    end_effector_bounds: dict[str, list[float]] = field( # bounds for the end-effector in x,y,z direction
+        default_factory=lambda: {
+            "min": [-1.0, -1.0, -1.0],  # min x, y, z
+            "max": [1.0, 1.0, 1.0],  # max x, y, z
+        }
+    )
+
+    max_gripper_pos: float = 50 # maximum gripper position that the gripper will be open at
+
+    end_effector_step_sizes: dict[str, float] = field( # maximum step size for the end-effector in x,y,z direction
+        default_factory=lambda: {
+            "x": 0.02,
+            "y": 0.02,
+            "z": 0.02,
+        }
+    )
+```
+
+The `Teleoperator` defines the teleoperation device. You can check the list of available teleoperators in `lerobot/teleoperators`.
+
+**Setting up the Gamepad**
+
+The gamepad provides a very convenient way to control the robot and the episode state.
+
+To setup the gamepad, you need to set the `control_mode` to `"gamepad"` and define the `teleop` section in the configuration file.
+
+```json
+    "teleop": {
+        "type": "gamepad",
+        "use_gripper": true
+    },
+```
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/gamepad_guide.jpg?raw=true" alt="Figure shows the control mappings on a Logitech gamepad." title="Gamepad Control Mapping" width="100%"></img>
+</p>
+<p align="center"><i>Gamepad button mapping for robot control and episode management</i></p>
+
+**Setting up the SO101 leader**
+
+The SO101 leader arm has reduced gears that allows it to move and track the follower arm during exploration. Therefore, taking over is much smoother than the gearless SO100.
+
+To setup the SO101 leader, you need to set the `control_mode` to `"leader"` and define the `teleop` section in the configuration file.
+
+```json
+    "teleop": {
+        "type": "so101_leader",
+        "port": "/dev/tty.usbmodem585A0077921", # check your port number
+        "use_degrees": true
+    },
+```
+
+In order to annotate the success/failure of the episode, **you will need** to use a keyboard to press `s` for success, `esc` for failure.
+During the online training, press `space` to take over the policy and `space` again to give the control back to the policy.
+
+<details>
+<summary><strong>Video: SO101 leader teleoperation</strong></summary>
+
+<div class="video-container">
+  <video controls width="600">
+    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/so101_leader_tutorial.mp4" type="video/mp4" />
+  </video>
+</div>
+
+<p align="center"><i>SO101 leader teleoperation example, the leader tracks the follower, press `space` to intervene</i></p>
+</details>
+
+**Recording Demonstrations**
+
+Start the recording process, an example of the config file can be found [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/env_config_so100.json):
+
+```bash
+python -m lerobot.scripts.rl.gym_manipulator --config_path src/lerobot/configs/env_config_so100.json
+```
+
+During recording:
+1. The robot will reset to the initial position defined in the configuration file `fixed_reset_joint_positions`
+2. Complete the task successfully
+3. The episode ends with a reward of 1 when you press the "success" button
+4. If the time limit is reached, or the fail button is pressed, the episode ends with a reward of 0
+5. You can rerecord an episode by pressing the "rerecord" button
+6. The process automatically continues to the next episode
+7. After recording all episodes, the dataset is pushed to the Hugging Face Hub (optional) and saved locally
+
+
+### Processing the Dataset
+
+After collecting demonstrations, process them to determine optimal camera crops.
+Reinforcement learning is sensitive to background distractions, so it is important to crop the images to the relevant workspace area.
+
+Visual RL algorithms learn directly from pixel inputs, making them vulnerable to irrelevant visual information. Background elements like changing lighting, shadows, people moving, or objects outside the workspace can confuse the learning process. Good ROI selection should:
+- Include only the essential workspace where the task happens
+- Capture the robot's end-effector and all objects involved in the task
+- Exclude unnecessary background elements and distractions
+
+Note: If you already know the crop parameters, you can skip this step and just set the `crop_params_dict` in the configuration file during recording.
+
+**Determining Crop Parameters**
+
+Use the `crop_dataset_roi.py` script to interactively select regions of interest in your camera images:
+
+```bash
+python -m lerobot.scripts.rl.crop_dataset_roi --repo-id username/pick_lift_cube
+```
+
+1. For each camera view, the script will display the first frame
+2. Draw a rectangle around the relevant workspace area
+3. Press 'c' to confirm the selection
+4. Repeat for all camera views
+5. The script outputs cropping parameters and creates a new cropped dataset
+
+Example output:
+```
+Selected Rectangular Regions of Interest (top, left, height, width):
+observation.images.side: [180, 207, 180, 200]
+observation.images.front: [180, 250, 120, 150]
+```
+
+<p align="center">
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/crop_dataset.gif" width="600"/>
+</p>
+
+<p align="center"><i>Interactive cropping tool for selecting regions of interest</i></p>
+
+
+**Updating Configuration**
+
+Add these crop parameters to your training configuration:
+
+```json
+"crop_params_dict": {
+    "observation.images.side": [180, 207, 180, 200],
+    "observation.images.front": [180, 250, 120, 150]
+},
+"resize_size": [128, 128]
+```
+
+**Recommended image resolution**
+
+Most vision-based policies have been validated on square inputs of either **128×128** (default) or **64×64** pixels.  We therefore advise setting the resize_size parameter to [128, 128] – or [64, 64] if you need to save GPU memory and bandwidth.  Other resolutions are possible but have not been extensively tested.
+
+
+### Training a Reward Classifier
+
+The reward classifier plays an important role in the HIL-SERL workflow by automating reward assignment and automatically detecting episode success. Instead of manually defining reward functions or relying on human feedback for every timestep, the reward classifier learns to predict success/failure from visual observations. This enables the RL algorithm to learn efficiently by providing consistent and automated reward signals based on the robot's camera inputs.
+
+This guide explains how to train a reward classifier for human-in-the-loop reinforcement learning implementation of LeRobot. Reward classifiers learn to predict the reward value given a state which can be used in an RL setup to train a policy.
+
+**Note**: Training a reward classifier is optional. You can start the first round of RL experiments by annotating the success manually with your gamepad or keyboard device.
+
+The reward classifier implementation in `modeling_classifier.py` uses a pretrained vision model to process the images. It can output either a single value for binary rewards to predict success/fail cases or multiple values for multi-class settings.
+
+**Collecting a Dataset for the reward classifier**
+
+Before training, you need to collect a dataset with labeled examples. The `record_dataset` function in `gym_manipulator.py` enables the process of collecting a dataset of observations, actions, and rewards.
+
+To collect a dataset, you need to modify some parameters in the environment configuration based on HILSerlRobotEnvConfig.
+
+```bash
+python -m lerobot.scripts.rl.gym_manipulator --config_path src/lerobot/configs/reward_classifier_train_config.json
+```
+
+**Key Parameters for Data Collection**
+
+- **mode**: set it to `"record"` to collect a dataset
+- **repo_id**: `"hf_username/dataset_name"`, name of the dataset and repo on the hub
+- **num_episodes**: Number of episodes to record
+- **number_of_steps_after_success**: Number of additional frames to record after a success (reward=1) is detected
+- **fps**: Number of frames per second to record
+- **push_to_hub**: Whether to push the dataset to the hub
+
+The `number_of_steps_after_success` parameter is crucial as it allows you to collect more positive examples. When a success is detected, the system will continue recording for the specified number of steps while maintaining the reward=1 label. Otherwise, there won't be enough states in the dataset labeled to 1 to train a good classifier.
+
+Example configuration section for data collection:
+
+```json
+{
+    "mode": "record",
+    "repo_id": "hf_username/dataset_name",
+    "dataset_root": "data/your_dataset",
+    "num_episodes": 20,
+    "push_to_hub": true,
+    "fps": 10,
+    "number_of_steps_after_success": 15
+}
+```
+
+**Reward Classifier Configuration**
+
+The reward classifier is configured using `configuration_classifier.py`. Here are the key parameters:
+
+- **model_name**: Base model architecture (e.g., we mainly use `"helper2424/resnet10"`)
+- **model_type**: `"cnn"` or `"transformer"`
+- **num_cameras**: Number of camera inputs
+- **num_classes**: Number of output classes (typically 2 for binary success/failure)
+- **hidden_dim**: Size of hidden representation
+- **dropout_rate**: Regularization parameter
+- **learning_rate**: Learning rate for optimizer
+
+Example configuration for training the [reward classifier](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/reward_classifier_train_config.json):
+
+```json
+{
+  "policy": {
+    "type": "reward_classifier",
+    "model_name": "helper2424/resnet10",
+    "model_type": "cnn",
+    "num_cameras": 2,
+    "num_classes": 2,
+    "hidden_dim": 256,
+    "dropout_rate": 0.1,
+    "learning_rate": 1e-4,
+    "device": "cuda",
+    "use_amp": true,
+    "input_features": {
+      "observation.images.front": {
+        "type": "VISUAL",
+        "shape": [3, 128, 128]
+      },
+      "observation.images.side": {
+        "type": "VISUAL",
+        "shape": [3, 128, 128]
+      }
+    }
+  }
+}
+```
+
+**Training the Classifier**
+
+To train the classifier, use the `train.py` script with your configuration:
+
+```bash
+python -m lerobot.scripts.train --config_path path/to/reward_classifier_train_config.json
+```
+
+**Deploying and Testing the Model**
+
+To use your trained reward classifier, configure the `HILSerlRobotEnvConfig` to use your model:
+
+```python
+env_config = HILSerlRobotEnvConfig(
+    reward_classifier_pretrained_path="path_to_your_pretrained_trained_model",
+    # Other environment parameters
+)
+```
+or set the argument in the json config file.
+
+```json
+{
+    "reward_classifier_pretrained_path": "path_to_your_pretrained_model"
+}
+```
+
+Run `gym_manipulator.py` to test the model.
+```bash
+python -m lerobot.scripts.rl.gym_manipulator --config_path path/to/env_config.json
+```
+
+The reward classifier will automatically provide rewards based on the visual input from the robot's cameras.
+
+**Example Workflow for training the reward classifier**
+
+1. **Create the configuration files**:
+   Create the necessary json configuration files for the reward classifier and the environment. Check the examples [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/tree/main).
+
+2. **Collect a dataset**:
+   ```bash
+   python -m lerobot.scripts.rl.gym_manipulator --config_path src/lerobot/configs/env_config.json
+   ```
+
+3. **Train the classifier**:
+   ```bash
+   python -m lerobot.scripts.train --config_path src/lerobot/configs/reward_classifier_train_config.json
+   ```
+
+4. **Test the classifier**:
+   ```bash
+   python -m lerobot.scripts.rl.gym_manipulator --config_path src/lerobot/configs/env_config.json
+   ```
+
+### Training with Actor-Learner
+
+The LeRobot system uses a distributed actor-learner architecture for training. This architecture decouples robot interactions from the learning process, allowing them to run concurrently without blocking each other. The actor server handles robot observations and actions, sending interaction data to the learner server. The learner server performs gradient descent and periodically updates the actor's policy weights. You will need to start two processes: a learner and an actor.
+
+**Configuration Setup**
+
+Create a training configuration file (example available [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/train_config_hilserl_so100.json)). The training config is based on the main `TrainRLServerPipelineConfig` class in `lerobot/configs/train.py`.
+
+1. Configure the policy settings (`type="sac"`, `device`, etc.)
+2. Set `dataset` to your cropped dataset
+3. Configure environment settings with crop parameters
+4. Check the other parameters related to SAC in [configuration_sac.py](https://github.com/huggingface/lerobot/blob/19bb621a7d0a31c20cd3cc08b1dbab68d3031454/lerobot/policies/sac/configuration_sac.py#L79).
+5. Verify that the `policy` config is correct with the right `input_features` and `output_features` for your task.
+
+**Starting the Learner**
+
+First, start the learner server process:
+
+```bash
+python -m lerobot.scripts.rl.learner --config_path src/lerobot/configs/train_config_hilserl_so100.json
+```
+
+The learner:
+- Initializes the policy network
+- Prepares replay buffers
+- Opens a `gRPC` server to communicate with actors
+- Processes transitions and updates the policy
+
+**Starting the Actor**
+
+In a separate terminal, start the actor process with the same configuration:
+
+```bash
+python -m lerobot.scripts.rl.actor --config_path src/lerobot/configs/train_config_hilserl_so100.json
+```
+
+The actor:
+- Connects to the learner via `gRPC`
+- Initializes the environment
+- Execute rollouts of the policy to collect experience
+- Sends transitions to the learner
+- Receives updated policy parameters
+
+**Training Flow**
+
+The training proceeds automatically:
+
+1. The actor executes the policy in the environment
+2. Transitions are collected and sent to the learner
+3. The learner updates the policy based on these transitions
+4. Updated policy parameters are sent back to the actor
+5. The process continues until the specified step limit is reached
+
+**Human in the Loop**
+
+- The key to learning efficiently is to have human interventions to provide corrective feedback and completing the task to aide the policy learning and exploration.
+- To perform human interventions, you can press the upper right trigger button on the gamepad (or the `space` key on the keyboard). This will pause the policy actions and allow you to take over.
+- A successful experiment is one where the human has to intervene at the start but then reduces the amount of interventions as the policy improves. You can monitor the intervention rate in the `wandb` dashboard.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/hil_effect.png?raw=true" alt="Figure shows the control mappings on a Logitech gamepad." title="Gamepad Control Mapping" width="100%"></img>
+</p>
+
+<p align="center"><i>Example showing how human interventions help guide policy learning over time</i></p>
+
+- The figure shows the plot of the episodic reward over interaction step. The figure shows the effect of human interventions on the policy learning.
+- The orange curve is an experiment without any human interventions. While the pink and blue curves are experiments with human interventions.
+- We can observe that the number of steps where the policy starts achieving the maximum reward is cut by a quarter when human interventions are present.
+
+**Monitoring and Debugging**
+
+If you have `wandb.enable` set to `true` in your configuration, you can monitor training progress in real-time through the [Weights & Biases](https://wandb.ai/site/) dashboard.
+
+### Guide to Human Interventions
+The learning process is very sensitive to the intervention strategy. It will takes a few runs to understand how to intervene effectively. Some tips and hints:
+- Allow the policy to explore for a few episodes at the start of training.
+- Avoid intervening for long periods of time. Try to intervene in situation to correct the robot's behaviour when it goes off track.
+- Once the policy starts achieving the task, even if its not perfect, you can limit your interventions to simple quick actions like a simple grasping commands.
+
+The ideal behaviour is that your intervention rate should drop gradually during training as shown in the figure below.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/intervention_rate_tutorial_rl.png?raw=true" alt="Intervention rate" title="Intervention rate during training" width="100%"></img>
+</p>
+
+<p align="center"><i>Plot of the intervention rate during a training run on a pick and lift cube task</i></p>
+
+### Key hyperparameters to tune
+
+Some configuration values have a disproportionate impact on training stability and speed:
+
+- **`temperature_init`** (`policy.temperature_init`) – initial entropy temperature in SAC. Higher values encourage more exploration; lower values make the policy more deterministic early on. A good starting point is `1e-2`. We observed that setting it too high can make human interventions ineffective and slow down learning.
+- **`policy_parameters_push_frequency`** (`policy.actor_learner_config.policy_parameters_push_frequency`) – interval in *seconds* between two weight pushes from the learner to the actor. The default is `4 s`. Decrease to **1-2 s** to provide fresher weights (at the cost of more network traffic); increase only if your connection is slow, as this will reduce sample efficiency.
+- **`storage_device`** (`policy.storage_device`) – device on which the learner keeps the policy parameters. If you have spare GPU memory, set this to `"cuda"` (instead of the default `"cpu"`). Keeping the weights on-GPU removes CPU→GPU transfer overhead and can significantly increase the number of learner updates per second.
+
+
+Congrats 🎉, you have finished this tutorial!
+
+> [!TIP]
+>  If you have any questions or need help, please reach out on [Discord](https://discord.com/invite/s3KuuzsPFb).
+
+Paper citation:
+```
+@article{luo2024precise,
+  title={Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning},
+  author={Luo, Jianlan and Xu, Charles and Wu, Jeffrey and Levine, Sergey},
+  journal={arXiv preprint arXiv:2410.21845},
+  year={2024}
+}
+```

docs/source/hilserl_sim.mdx

@@ -0,0 +1,120 @@
+# Train RL in Simulation
+
+This guide explains how to use the `gym_hil` simulation environments as an alternative to real robots when working with the LeRobot framework for Human-In-the-Loop (HIL) reinforcement learning.
+
+`gym_hil` is a package that provides Gymnasium-compatible simulation environments specifically designed for Human-In-the-Loop reinforcement learning. These environments allow you to:
+
+- Train policies in simulation to test the RL stack before training on real robots
+
+- Collect demonstrations in sim using external devices like gamepads or keyboards
+- Perform human interventions during policy learning
+
+Currently, the main environment is a Franka Panda robot simulation based on MuJoCo, with tasks like picking up a cube.
+
+
+## Installation
+
+First, install the `gym_hil` package within the LeRobot environment:
+
+```bash
+pip install -e ".[hilserl]"
+```
+
+## What do I need?
+
+- A gamepad or keyboard to control the robot
+- A Nvidia GPU
+
+
+
+## Configuration
+
+To use `gym_hil` with LeRobot, you need to create a configuration file. An example is provided [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/gym_hil_env.json). Key configuration sections include:
+
+### Environment Type and Task
+
+```json
+{
+    "type": "hil",
+    "name": "franka_sim",
+    "task": "PandaPickCubeGamepad-v0",
+    "device": "cuda"
+}
+```
+
+Available tasks:
+- `PandaPickCubeBase-v0`: Basic environment
+- `PandaPickCubeGamepad-v0`: With gamepad control
+- `PandaPickCubeKeyboard-v0`: With keyboard control
+
+### Gym Wrappers Configuration
+
+```json
+"wrapper": {
+    "gripper_penalty": -0.02,
+    "control_time_s": 15.0,
+    "use_gripper": true,
+    "fixed_reset_joint_positions": [0.0, 0.195, 0.0, -2.43, 0.0, 2.62, 0.785],
+    "end_effector_step_sizes": {
+        "x": 0.025,
+        "y": 0.025,
+        "z": 0.025
+    },
+    "control_mode": "gamepad"
+    }
+```
+
+Important parameters:
+- `gripper_penalty`: Penalty for excessive gripper movement
+- `use_gripper`: Whether to enable gripper control
+- `end_effector_step_sizes`: Size of the steps in the x,y,z axes of the end-effector
+- `control_mode`: Set to `"gamepad"` to use a gamepad controller
+
+## Running with HIL RL of LeRobot
+
+### Basic Usage
+
+To run the environment, set mode to null:
+
+```python
+python -m lerobot.scripts.rl.gym_manipulator --config_path path/to/gym_hil_env.json
+```
+
+### Recording a Dataset
+
+To collect a dataset, set the mode to `record` whilst defining the repo_id and number of episodes to record:
+
+```python
+python -m lerobot.scripts.rl.gym_manipulator --config_path path/to/gym_hil_env.json
+```
+
+### Training a Policy
+
+To train a policy, checkout the configuration example available [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/train_gym_hil_env.json) and run the actor and learner servers:
+
+```python
+python -m lerobot.scripts.rl.actor --config_path path/to/train_gym_hil_env.json
+```
+
+In a different terminal, run the learner server:
+
+```python
+python -m lerobot.scripts.rl.learner --config_path path/to/train_gym_hil_env.json
+```
+
+The simulation environment provides a safe and repeatable way to develop and test your Human-In-the-Loop reinforcement learning components before deploying to real robots.
+
+Congrats 🎉, you have finished this tutorial!
+
+> [!TIP]
+>  If you have any questions or need help, please reach out on [Discord](https://discord.com/invite/s3KuuzsPFb).
+
+Paper citation:
+```
+@article{luo2024precise,
+  title={Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning},
+  author={Luo, Jianlan and Xu, Charles and Wu, Jeffrey and Levine, Sergey},
+  journal={arXiv preprint arXiv:2410.21845},
+  year={2024}
+}
+```

docs/source/hope_jr.mdx

@@ -0,0 +1 @@
+../../src/lerobot/robots/hope_jr/hope_jr.mdx

docs/source/il_robots.mdx

@@ -0,0 +1,547 @@
+# Imitation Learning on Real-World Robots
+
+This tutorial will explain how to train a neural network to control a real robot autonomously.
+
+**You'll learn:**
+1. How to record and visualize your dataset.
+2. How to train a policy using your data and prepare it for evaluation.
+3. How to evaluate your policy and visualize the results.
+
+By following these steps, you'll be able to replicate tasks, such as picking up a Lego block and placing it in a bin with a high success rate, as shown in the video below.
+
+<details>
+<summary><strong>Video: pickup lego block task</strong></summary>
+
+<div class="video-container">
+  <video controls width="600">
+    <source src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/lerobot_task.mp4" type="video/mp4" />
+  </video>
+</div>
+
+</details>
+
+This tutorial isn’t tied to a specific robot: we walk you through the commands and API snippets you can adapt for any supported platform.
+
+During data collection, you’ll use a “teloperation” device, such as a leader arm or keyboard to teleoperate the robot and record its motion trajectories.
+
+Once you’ve gathered enough trajectories, you’ll train a neural network to imitate these trajectories and deploy the trained model so your robot can perform the task autonomously.
+
+If you run into any issues at any point, jump into our [Discord community](https://discord.com/invite/s3KuuzsPFb) for support.
+
+## Set up and Calibrate
+
+If you haven't yet set up and calibrated your robot and teleop device, please do so by following the robot-specific tutorial.
+
+## Teleoperate
+
+In this example, we’ll demonstrate how to teleoperate the SO101 robot. For each command, we also provide a corresponding API example.
+
+Note that the `id` associated with a robot is used to store the calibration file. It's important to use the same `id` when teleoperating, recording, and evaluating when using the same setup.
+
+<hfoptions id="teleoperate_so101">
+<hfoption id="Command">
+```bash
+python -m lerobot.teleoperate \
+    --robot.type=so101_follower \
+    --robot.port=/dev/tty.usbmodem58760431541 \
+    --robot.id=my_awesome_follower_arm \
+    --teleop.type=so101_leader \
+    --teleop.port=/dev/tty.usbmodem58760431551 \
+    --teleop.id=my_awesome_leader_arm
+```
+</hfoption>
+<hfoption id="API example">
+```python
+from lerobot.teleoperators.so101_leader import SO101LeaderConfig, SO101Leader
+from lerobot.robots.so101_follower import SO101FollowerConfig, SO101Follower
+
+robot_config = SO101FollowerConfig(
+    port="/dev/tty.usbmodem58760431541",
+    id="my_red_robot_arm",
+)
+
+teleop_config = SO101LeaderConfig(
+    port="/dev/tty.usbmodem58760431551",
+    id="my_blue_leader_arm",
+)
+
+robot = SO101Follower(robot_config)
+teleop_device = SO101Leader(teleop_config)
+robot.connect()
+teleop_device.connect()
+
+while True:
+    action = teleop_device.get_action()
+    robot.send_action(action)
+```
+</hfoption>
+</hfoptions>
+
+The teleoperate command will automatically:
+1. Identify any missing calibrations and initiate the calibration procedure.
+2. Connect the robot and teleop device and start teleoperation.
+
+## Cameras
+
+To add cameras to your setup, follow this [Guide](./cameras#setup-cameras).
+
+## Teleoperate with cameras
+
+With `rerun`, you can teleoperate again while simultaneously visualizing the camera feeds and joint positions. In this example, we’re using the Koch arm.
+
+<hfoptions id="teleoperate_koch_camera">
+<hfoption id="Command">
+```bash
+python -m lerobot.teleoperate \
+    --robot.type=koch_follower \
+    --robot.port=/dev/tty.usbmodem58760431541 \
+    --robot.id=my_awesome_follower_arm \
+    --robot.cameras="{ front: {type: opencv, index_or_path: 0, width: 1920, height: 1080, fps: 30}}" \
+    --teleop.type=koch_leader \
+    --teleop.port=/dev/tty.usbmodem58760431551 \
+    --teleop.id=my_awesome_leader_arm \
+    --display_data=true
+```
+</hfoption>
+<hfoption id="API example">
+```python
+from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.teleoperators.koch_leader import KochLeaderConfig, KochLeader
+from lerobot.robots.koch_follower import KochFollowerConfig, KochFollower
+
+camera_config = {
+    "front": OpenCVCameraConfig(index_or_path=0, width=1920, height=1080, fps=30)
+}
+
+robot_config = KochFollowerConfig(
+    port="/dev/tty.usbmodem585A0076841",
+    id="my_red_robot_arm",
+    cameras=camera_config
+)
+
+teleop_config = KochLeaderConfig(
+    port="/dev/tty.usbmodem58760431551",
+    id="my_blue_leader_arm",
+)
+
+robot = KochFollower(robot_config)
+teleop_device = KochLeader(teleop_config)
+robot.connect()
+teleop_device.connect()
+
+while True:
+    observation = robot.get_observation()
+    action = teleop_device.get_action()
+    robot.send_action(action)
+```
+</hfoption>
+</hfoptions>
+
+## Record a dataset
+
+Once you're familiar with teleoperation, you can record your first dataset.
+
+We use the Hugging Face hub features for uploading your dataset. If you haven't previously used the Hub, make sure you can login via the cli using a write-access token, this token can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens).
+
+Add your token to the CLI by running this command:
+```bash
+huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
+```
+
+Then store your Hugging Face repository name in a variable:
+```bash
+HF_USER=$(huggingface-cli whoami | head -n 1)
+echo $HF_USER
+```
+
+Now you can record a dataset. To record 5 episodes and upload your dataset to the hub, adapt the code below for your robot and execute the command or API example.
+
+<hfoptions id="record">
+<hfoption id="Command">
+```bash
+python -m lerobot.record \
+    --robot.type=so101_follower \
+    --robot.port=/dev/tty.usbmodem585A0076841 \
+    --robot.id=my_awesome_follower_arm \
+    --robot.cameras="{ front: {type: opencv, index_or_path: 0, width: 1920, height: 1080, fps: 30}}" \
+    --teleop.type=so101_leader \
+    --teleop.port=/dev/tty.usbmodem58760431551 \
+    --teleop.id=my_awesome_leader_arm \
+    --display_data=true \
+    --dataset.repo_id=${HF_USER}/record-test \
+    --dataset.num_episodes=5 \
+    --dataset.single_task="Grab the black cube"
+```
+</hfoption>
+<hfoption id="API example">
+```python
+from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.utils import hw_to_dataset_features
+from lerobot.robots.so100_follower import SO100Follower, SO100FollowerConfig
+from lerobot.teleoperators.so100_leader.config_so100_leader import SO100LeaderConfig
+from lerobot.teleoperators.so100_leader.so100_leader import SO100Leader
+from lerobot.utils.control_utils import init_keyboard_listener
+from lerobot.utils.utils import log_say
+from lerobot.utils.visualization_utils import _init_rerun
+from lerobot.record import record_loop
+
+NUM_EPISODES = 5
+FPS = 30
+EPISODE_TIME_SEC = 60
+RESET_TIME_SEC = 10
+TASK_DESCRIPTION = "My task description"
+
+# Create the robot and teleoperator configurations
+camera_config = {"front": OpenCVCameraConfig(index_or_path=0, width=640, height=480, fps=FPS)}
+robot_config = SO100FollowerConfig(
+    port="/dev/tty.usbmodem58760434471", id="my_awesome_follower_arm", cameras=camera_config
+)
+teleop_config = SO100LeaderConfig(port="/dev/tty.usbmodem585A0077581", id="my_awesome_leader_arm")
+
+# Initialize the robot and teleoperator
+robot = SO100Follower(robot_config)
+teleop = SO100Leader(teleop_config)
+
+# Configure the dataset features
+action_features = hw_to_dataset_features(robot.action_features, "action")
+obs_features = hw_to_dataset_features(robot.observation_features, "observation")
+dataset_features = {**action_features, **obs_features}
+
+# Create the dataset
+dataset = LeRobotDataset.create(
+    repo_id="<hf_username>/<dataset_repo_id>",
+    fps=FPS,
+    features=dataset_features,
+    robot_type=robot.name,
+    use_videos=True,
+    image_writer_threads=4,
+)
+
+# Initialize the keyboard listener and rerun visualization
+_, events = init_keyboard_listener()
+_init_rerun(session_name="recording")
+
+# Connect the robot and teleoperator
+robot.connect()
+teleop.connect()
+
+episode_idx = 0
+while episode_idx < NUM_EPISODES and not events["stop_recording"]:
+    log_say(f"Recording episode {episode_idx + 1} of {NUM_EPISODES}")
+
+    record_loop(
+        robot=robot,
+        events=events,
+        fps=FPS,
+        teleop=teleop,
+        dataset=dataset,
+        control_time_s=EPISODE_TIME_SEC,
+        single_task=TASK_DESCRIPTION,
+        display_data=True,
+    )
+
+    # Reset the environment if not stopping or re-recording
+    if not events["stop_recording"] and (episode_idx < NUM_EPISODES - 1 or events["rerecord_episode"]):
+        log_say("Reset the environment")
+        record_loop(
+            robot=robot,
+            events=events,
+            fps=FPS,
+            teleop=teleop,
+            control_time_s=RESET_TIME_SEC,
+            single_task=TASK_DESCRIPTION,
+            display_data=True,
+        )
+
+    if events["rerecord_episode"]:
+        log_say("Re-recording episode")
+        events["rerecord_episode"] = False
+        events["exit_early"] = False
+        dataset.clear_episode_buffer()
+        continue
+
+    dataset.save_episode()
+    episode_idx += 1
+
+# Clean up
+log_say("Stop recording")
+robot.disconnect()
+teleop.disconnect()
+dataset.push_to_hub()
+```
+</hfoption>
+</hfoptions>
+
+#### Dataset upload
+Locally, your dataset is stored in this folder: `~/.cache/huggingface/lerobot/{repo-id}`. At the end of data recording, your dataset will be uploaded on your Hugging Face page (e.g. https://huggingface.co/datasets/cadene/so101_test) that you can obtain by running:
+```bash
+echo https://huggingface.co/datasets/${HF_USER}/so101_test
+```
+Your dataset will be automatically tagged with `LeRobot` for the community to find it easily, and you can also add custom tags (in this case `tutorial` for example).
+
+You can look for other LeRobot datasets on the hub by searching for `LeRobot` [tags](https://huggingface.co/datasets?other=LeRobot).
+
+You can also push your local dataset to the Hub manually, running:
+```bash
+huggingface-cli upload ${HF_USER}/record-test ~/.cache/huggingface/lerobot/{repo-id} --repo-type dataset
+```
+
+
+#### Record function
+
+The `record` function provides a suite of tools for capturing and managing data during robot operation:
+
+##### 1. Data Storage
+- Data is stored using the `LeRobotDataset` format and is stored on disk during recording.
+- By default, the dataset is pushed to your Hugging Face page after recording.
+  - To disable uploading, use `--dataset.push_to_hub=False`.
+
+##### 2. Checkpointing and Resuming
+- Checkpoints are automatically created during recording.
+- If an issue occurs, you can resume by re-running the same command with `--resume=true`.
+- To start recording from scratch, **manually delete** the dataset directory.
+
+##### 3. Recording Parameters
+Set the flow of data recording using command-line arguments:
+- `--dataset.episode_time_s=60`
+  Duration of each data recording episode (default: **60 seconds**).
+- `--dataset.reset_time_s=60`
+  Duration for resetting the environment after each episode (default: **60 seconds**).
+- `--dataset.num_episodes=50`
+  Total number of episodes to record (default: **50**).
+
+##### 4. Keyboard Controls During Recording
+Control the data recording flow using keyboard shortcuts:
+- Press **Right Arrow (`→`)**: Early stop the current episode or reset time and move to the next.
+- Press **Left Arrow (`←`)**: Cancel the current episode and re-record it.
+- Press **Escape (`ESC`)**: Immediately stop the session, encode videos, and upload the dataset.
+
+#### Tips for gathering data
+
+Once you're comfortable with data recording, you can create a larger dataset for training. A good starting task is grasping an object at different locations and placing it in a bin. We suggest recording at least 50 episodes, with 10 episodes per location. Keep the cameras fixed and maintain consistent grasping behavior throughout the recordings. Also make sure the object you are manipulating is visible on the camera's. A good rule of thumb is you should be able to do the task yourself by only looking at the camera images.
+
+In the following sections, you’ll train your neural network. After achieving reliable grasping performance, you can start introducing more variations during data collection, such as additional grasp locations, different grasping techniques, and altering camera positions.
+
+Avoid adding too much variation too quickly, as it may hinder your results.
+
+If you want to dive deeper into this important topic, you can check out the [blog post](https://huggingface.co/blog/lerobot-datasets#what-makes-a-good-dataset) we wrote on what makes a good dataset.
+
+
+#### Troubleshooting:
+- On Linux, if the left and right arrow keys and escape key don't have any effect during data recording, make sure you've set the `$DISPLAY` environment variable. See [pynput limitations](https://pynput.readthedocs.io/en/latest/limitations.html#linux).
+
+## Visualize a dataset
+
+If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:
+```bash
+echo ${HF_USER}/so101_test
+```
+
+## Replay an episode
+
+A useful feature is the `replay` function, which allows you to replay any episode that you've recorded or episodes from any dataset out there. This function helps you test the repeatability of your robot's actions and assess transferability across robots of the same model.
+
+You can replay the first episode on your robot with either the command below or with the API example:
+
+<hfoptions id="replay">
+<hfoption id="Command">
+```bash
+python -m lerobot.replay \
+    --robot.type=so101_follower \
+    --robot.port=/dev/tty.usbmodem58760431541 \
+    --robot.id=my_awesome_follower_arm \
+    --dataset.repo_id=${HF_USER}/record-test \
+    --dataset.episode=0 # choose the episode you want to replay
+```
+</hfoption>
+<hfoption id="API example">
+```python
+import time
+
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.robots.so100_follower.config_so100_follower import SO100FollowerConfig
+from lerobot.robots.so100_follower.so100_follower import SO100Follower
+from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.utils import log_say
+
+episode_idx = 0
+
+robot_config = SO100FollowerConfig(port="/dev/tty.usbmodem58760434471", id="my_awesome_follower_arm")
+
+robot = SO100Follower(robot_config)
+robot.connect()
+
+dataset = LeRobotDataset("<hf_username>/<dataset_repo_id>", episodes=[episode_idx])
+actions = dataset.hf_dataset.select_columns("action")
+
+log_say(f"Replaying episode {episode_idx}")
+for idx in range(dataset.num_frames):
+    t0 = time.perf_counter()
+
+    action = {
+        name: float(actions[idx]["action"][i]) for i, name in enumerate(dataset.features["action"]["names"])
+    }
+    robot.send_action(action)
+
+    busy_wait(1.0 / dataset.fps - (time.perf_counter() - t0))
+
+robot.disconnect()
+```
+</hfoption>
+</hfoptions>
+
+Your robot should replicate movements similar to those you recorded. For example, check out [this video](https://x.com/RemiCadene/status/1793654950905680090) where we use `replay` on a Aloha robot from [Trossen Robotics](https://www.trossenrobotics.com).
+
+## Train a policy
+
+To train a policy to control your robot, use the [`python -m lerobot.scripts.train`](../src/lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
+```bash
+python -m lerobot.scripts.train \
+  --dataset.repo_id=${HF_USER}/so101_test \
+  --policy.type=act \
+  --output_dir=outputs/train/act_so101_test \
+  --job_name=act_so101_test \
+  --policy.device=cuda \
+  --wandb.enable=true \
+  --policy.repo_id=${HF_USER}/my_policy
+```
+
+Let's explain the command:
+1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/so101_test`.
+2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../src/lerobot/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
+4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
+5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
+
+Training should take several hours. You will find checkpoints in `outputs/train/act_so101_test/checkpoints`.
+
+To resume training from a checkpoint, below is an example command to resume from `last` checkpoint of the `act_so101_test` policy:
+```bash
+python -m lerobot.scripts.train \
+  --config_path=outputs/train/act_so101_test/checkpoints/last/pretrained_model/train_config.json \
+  --resume=true
+```
+
+If you do not want to push your model to the hub after training use `--policy.push_to_hub=false`.
+
+Additionally you can provide extra `tags` or specify a `license` for your model or make the model repo `private` by adding this: `--policy.private=true --policy.tags=\[ppo,rl\] --policy.license=mit`
+
+#### Train using Collab
+If your local computer doesn't have a powerful GPU you could utilize Google Collab to train your model by following the [ACT training notebook](./notebooks#training-act).
+
+#### Upload policy checkpoints
+
+Once training is done, upload the latest checkpoint with:
+```bash
+huggingface-cli upload ${HF_USER}/act_so101_test \
+  outputs/train/act_so101_test/checkpoints/last/pretrained_model
+```
+
+You can also upload intermediate checkpoints with:
+```bash
+CKPT=010000
+huggingface-cli upload ${HF_USER}/act_so101_test${CKPT} \
+  outputs/train/act_so101_test/checkpoints/${CKPT}/pretrained_model
+```
+
+## Run inference and evaluate your policy
+
+You can use the `record` script from [`lerobot/record.py`](https://github.com/huggingface/lerobot/blob/main/lerobot/record.py) with a policy checkpoint as input, to run inference and evaluate your policy. For instance, run this command or API example to run inference and record 10 evaluation episodes:
+
+<hfoptions id="eval">
+<hfoption id="Command">
+```bash
+python -m lerobot.record  \
+  --robot.type=so100_follower \
+  --robot.port=/dev/ttyACM1 \
+  --robot.cameras="{ up: {type: opencv, index_or_path: /dev/video10, width: 640, height: 480, fps: 30}, side: {type: intelrealsense, serial_number_or_name: 233522074606, width: 640, height: 480, fps: 30}}" \
+  --robot.id=my_awesome_follower_arm \
+  --display_data=false \
+  --dataset.repo_id=${HF_USER}/eval_so100 \
+  --dataset.single_task="Put lego brick into the transparent box" \
+  # <- Teleop optional if you want to teleoperate in between episodes \
+  # --teleop.type=so100_leader \
+  # --teleop.port=/dev/ttyACM0 \
+  # --teleop.id=my_awesome_leader_arm \
+  --policy.path=${HF_USER}/my_policy
+```
+</hfoption>
+<hfoption id="API example">
+```python
+from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.utils import hw_to_dataset_features
+from lerobot.policies.act.modeling_act import ACTPolicy
+from lerobot.robots.so100_follower.config_so100_follower import SO100FollowerConfig
+from lerobot.robots.so100_follower.so100_follower import SO100Follower
+from lerobot.utils.control_utils import init_keyboard_listener
+from lerobot.utils.utils import log_say
+from lerobot.utils.visualization_utils import _init_rerun
+from lerobot.record import record_loop
+
+NUM_EPISODES = 5
+FPS = 30
+EPISODE_TIME_SEC = 60
+TASK_DESCRIPTION = "My task description"
+
+# Create the robot configuration
+camera_config = {"front": OpenCVCameraConfig(index_or_path=0, width=640, height=480, fps=FPS)}
+robot_config = SO100FollowerConfig(
+    port="/dev/tty.usbmodem58760434471", id="my_awesome_follower_arm", cameras=camera_config
+)
+
+# Initialize the robot
+robot = SO100Follower(robot_config)
+
+# Initialize the policy
+policy = ACTPolicy.from_pretrained("<hf_username>/<my_policy_repo_id>")
+
+# Configure the dataset features
+action_features = hw_to_dataset_features(robot.action_features, "action")
+obs_features = hw_to_dataset_features(robot.observation_features, "observation")
+dataset_features = {**action_features, **obs_features}
+
+# Create the dataset
+dataset = LeRobotDataset.create(
+    repo_id="<hf_username>/eval_<dataset_repo_id>",
+    fps=FPS,
+    features=dataset_features,
+    robot_type=robot.name,
+    use_videos=True,
+    image_writer_threads=4,
+)
+
+# Initialize the keyboard listener and rerun visualization
+_, events = init_keyboard_listener()
+_init_rerun(session_name="recording")
+
+# Connect the robot
+robot.connect()
+
+for episode_idx in range(NUM_EPISODES):
+    log_say(f"Running inference, recording eval episode {episode_idx + 1} of {NUM_EPISODES}")
+
+    # Run the policy inference loop
+    record_loop(
+        robot=robot,
+        events=events,
+        fps=FPS,
+        policy=policy,
+        dataset=dataset,
+        control_time_s=EPISODE_TIME_SEC,
+        single_task=TASK_DESCRIPTION,
+        display_data=True,
+    )
+
+    dataset.save_episode()
+
+# Clean up
+robot.disconnect()
+dataset.push_to_hub()
+```
+</hfoption>
+</hfoptions>
+
+As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
+1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_act_so101_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_so101_test`).
+2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_so101_test`).

docs/source/il_sim.mdx

@@ -0,0 +1,152 @@
+# Imitation Learning in Sim
+
+This tutorial will explain how to train a neural network to control a robot in simulation with imitation learning.
+
+**You'll learn:**
+1. How to record a dataset in simulation with [gym-hil](https://github.com/huggingface/gym-hil) and visualize the dataset.
+2. How to train a policy using your data.
+3. How to evaluate your policy in simulation and visualize the results.
+
+For the simulation environment we use the same [repo](https://github.com/huggingface/gym-hil) that is also being used by the Human-In-the-Loop (HIL) reinforcement learning algorithm.
+This environment is based on [MuJoCo](https://mujoco.org) and allows you to record datasets in LeRobotDataset format.
+Teleoperation is easiest with a controller like the Logitech F710, but you can also use your keyboard if you are up for the challenge.
+
+## Installation
+
+First, install the `gym_hil` package within the LeRobot environment, go to your LeRobot folder and run this command:
+
+```bash
+pip install -e ".[hilserl]"
+```
+
+## Teleoperate and Record a Dataset
+
+To use `gym_hil` with LeRobot, you need to use a configuration file. An example config file can be found [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/env_config_gym_hil_il.json).
+
+To teleoperate and collect a dataset, we need to modify this config file and you should add your `repo_id` here: `"repo_id": "il_gym",` and `"num_episodes": 30,` and make sure you set `mode` to `record`, "mode": "record".
+
+If you do not have a Nvidia GPU also change `"device": "cuda"` parameter in the config file (for example to `mps` for MacOS).
+
+By default the config file assumes you use a controller. To use your keyboard please change the envoirment specified at `"task"` in the config file and set it to `"PandaPickCubeKeyboard-v0"`.
+
+Then we can run this command to start:
+
+<hfoptions id="teleop_sim">
+<hfoption id="Linux">
+
+```bash
+python -m lerobot.scripts.rl.gym_manipulator --config_path path/to/env_config_gym_hil_il.json
+```
+
+</hfoption>
+<hfoption id="MacOS">
+
+```bash
+mjpython -m lerobot.scripts.rl.gym_manipulator --config_path path/to/env_config_gym_hil_il.json
+```
+
+</hfoption>
+</hfoptions>
+
+Once rendered you can teleoperate the robot with the gamepad or keyboard, below you can find the gamepad/keyboard controls.
+
+Note that to teleoperate the robot you have to hold the "Human Take Over Pause Policy" Button `RB` to enable control!
+
+**Gamepad Controls**
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/gamepad_guide.jpg?raw=true" alt="Figure shows the control mappings on a Logitech gamepad." title="Gamepad Control Mapping" width="100%"></img>
+</p>
+<p align="center"><i>Gamepad button mapping for robot control and episode management</i></p>
+
+**Keyboard controls**
+
+For keyboard controls use the `spacebar` to enable control and the following keys to move the robot:
+```bash
+  Arrow keys: Move in X-Y plane
+  Shift and Shift_R: Move in Z axis
+  Right Ctrl and Left Ctrl: Open and close gripper
+  ESC: Exit
+```
+
+## Visualize a dataset
+
+If you uploaded your dataset to the hub you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id.
+
+<p align="center">
+  <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/dataset_visualizer_sim.png" alt="Figure shows the dataset visualizer" title="Dataset visualization" width="100%"></img>
+</p>
+<p align="center"><i>Dataset visualizer</i></p>
+
+
+## Train a policy
+
+To train a policy to control your robot, use the [`python -m lerobot.scripts.train`](../src/lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
+```bash
+python -m lerobot.scripts.train \
+  --dataset.repo_id=${HF_USER}/il_gym \
+  --policy.type=act \
+  --output_dir=outputs/train/il_sim_test \
+  --job_name=il_sim_test \
+  --policy.device=cuda \
+  --wandb.enable=true
+```
+
+Let's explain the command:
+1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/il_gym`.
+2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../src/lerobot/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
+4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
+5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
+
+Training should take several hours, 100k steps (which is the default) will take about 1h on Nvidia A100. You will find checkpoints in `outputs/train/il_sim_test/checkpoints`.
+
+#### Train using Collab
+If your local computer doesn't have a powerful GPU you could utilize Google Collab to train your model by following the [ACT training notebook](./notebooks#training-act).
+
+#### Upload policy checkpoints
+
+Once training is done, upload the latest checkpoint with:
+```bash
+huggingface-cli upload ${HF_USER}/il_sim_test \
+  outputs/train/il_sim_test/checkpoints/last/pretrained_model
+```
+
+You can also upload intermediate checkpoints with:
+```bash
+CKPT=010000
+huggingface-cli upload ${HF_USER}/il_sim_test${CKPT} \
+  outputs/train/il_sim_test/checkpoints/${CKPT}/pretrained_model
+```
+
+## Evaluate your policy in Sim
+
+To evaluate your policy we have to use the config file that can be found [here](https://huggingface.co/datasets/aractingi/lerobot-example-config-files/blob/main/eval_config_gym_hil.json).
+
+Make sure to replace the `repo_id` with the dataset you trained on, for example `pepijn223/il_sim_dataset` and replace the `pretrained_policy_name_or_path` with your model id, for example `pepijn223/il_sim_model`
+
+Then you can run this command to visualize your trained policy
+
+<hfoptions id="eval_policy">
+<hfoption id="Linux">
+
+```bash
+python -m lerobot.scripts.rl.eval_policy --config_path=path/to/eval_config_gym_hil.json
+```
+
+</hfoption>
+<hfoption id="MacOS">
+
+```bash
+mjpython -m lerobot.scripts.rl.eval_policy --config_path=path/to/eval_config_gym_hil.json
+```
+
+</hfoption>
+</hfoptions>
+
+> [!WARNING]
+> While the main workflow of training ACT in simulation is straightforward, there is significant room for exploring  how to set up the task, define the initial state of the environment, and determine the type of data required during collection to learn the most effective policy. If your trained policy doesn't perform well, investigate the quality of the dataset it was trained on using our visualizers, as well as the action values and various hyperparameters related to ACT and the simulation.
+
+Congrats 🎉, you have finished this tutorial. If you want to continue with using LeRobot in simulation follow this [Tutorial on reinforcement learning in sim with HIL-SERL](https://huggingface.co/docs/lerobot/hilserl_sim)
+
+> [!TIP]
+>  If you have any questions or need help, please reach out on [Discord](https://discord.com/invite/s3KuuzsPFb).

docs/source/installation.mdx

@@ -2,6 +2,8 @@
 
 ## Install LeRobot
 
+Currently only available from source.
+
 Download our source code:
 ```bash
 git clone https://github.com/huggingface/lerobot.git
@@ -13,28 +15,6 @@ Create a virtual environment with Python 3.10, using [`Miniconda`](https://docs.
 conda create -y -n lerobot python=3.10
 ```
 
-Now restart the shell by running:
-<hfoptions id="shell_restart">
-<hfoption id="Windows">
-
-```bash
-source ~/.bashrc
-```
-</hfoption>
-<hfoption id="Mac">
-
-```bash
-source ~/.bash_profile
-```
-</hfoption>
-<hfoption id="zshell">
-
-```bash
-source ~/.zshrc
-```
-</hfoption>
-</hfoptions>
-
 Then activate your conda environment, you have to do this each time you open a shell to use lerobot:
 ```bash
 conda activate lerobot
@@ -51,14 +31,14 @@ conda install ffmpeg -c conda-forge
 >  ```bash
 >  conda install ffmpeg=7.1.1 -c conda-forge
 >  ```
->  - _[On Linux only]_ Install [ffmpeg build dependencies](https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#GettheDependencies) and [compile ffmpeg from source with libsvtav1](https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#libsvtav1), and make sure you use the corresponding ffmpeg binary to your install with `which ffmpeg`.
+>  - _[On Linux only]_ If you want to bring your own ffmpeg: Install [ffmpeg build dependencies](https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#GettheDependencies) and [compile ffmpeg from source with libsvtav1](https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#libsvtav1), and make sure you use the corresponding ffmpeg binary to your install with `which ffmpeg`.
 
 Install 🤗 LeRobot:
 ```bash
-cd lerobot && pip install ".[feetech]"
+pip install -e .
 ```
 
-## Troubleshooting
+### Troubleshooting
 If you encounter build errors, you may need to install additional dependencies: `cmake`, `build-essential`, and `ffmpeg libs`.
 To install these for linux run:
 ```bash
@@ -66,19 +46,27 @@ sudo apt-get install cmake build-essential python-dev pkg-config libavformat-dev
 ```
 For other systems, see: [Compiling PyAV](https://pyav.org/docs/develop/overview/installation.html#bring-your-own-ffmpeg)
 
-## Sim
-For simulations, 🤗 LeRobot comes with gymnasium environments that can be installed as extras:
-- [aloha](https://github.com/huggingface/gym-aloha)
-- [xarm](https://github.com/huggingface/gym-xarm)
-- [pusht](https://github.com/huggingface/gym-pusht)
+## Optional dependencies
 
-For instance, to install 🤗 LeRobot with aloha and pusht, use:
+LeRobot provides optional extras for specific functionalities. Multiple extras can be combined (e.g., `.[aloha,feetech]`). For all available extras, refer to `pyproject.toml`.
+
+### Simulations
+Install environment packages: `aloha` ([gym-aloha](https://github.com/huggingface/gym-aloha)), `xarm` ([gym-xarm](https://github.com/huggingface/gym-xarm)), or `pusht` ([gym-pusht](https://github.com/huggingface/gym-pusht))
+Example:
 ```bash
-pip install -e ".[aloha, pusht]"
+pip install -e ".[aloha]" # or "[pusht]" for example
 ```
 
-## W&B
+### Motor Control
+For Koch v1.1 install the Dynamixel SDK, for SO100/SO101/Moss install the Feetech SDK.
+```bash
+pip install -e ".[feetech]" # or "[dynamixel]" for example
+```
+
+### Experiment Tracking
 To use [Weights and Biases](https://docs.wandb.ai/quickstart) for experiment tracking, log in with
 ```bash
 wandb login
 ```
+
+You can now assemble your robot if it's not ready yet, look for your robot type on the left. Then follow the link below to use Lerobot with your robot.

docs/source/integrate_hardware.mdx

@@ -0,0 +1,318 @@
+# Bring Your Own Hardware
+
+This tutorial will explain how to integrate your own robot design into the LeRobot ecosystem and have it access all of our tools (data collection, control pipelines, policy training and inference).
+
+To that end, we provide the [`Robot`](https://github.com/huggingface/lerobot/blob/main/lerobot/robots/robot.py) base class in the LeRobot which specifies a standard interface for physical robot integration. Let's see how to implement it.
+
+## Prerequisites
+
+- Your own robot which exposes a communication interface (e.g. serial, CAN, TCP)
+- A way to read sensor data and send motor commands programmatically, e.g. manufacturer's SDK or API, or your own protocol implementation.
+- LeRobot installed in your environment. Follow our [Installation Guide](./installation).
+
+## Choose your motors
+
+If you're using Feetech or Dynamixel motors, LeRobot provides built-in bus interfaces:
+
+- [`FeetechMotorsBus`](https://github.com/huggingface/lerobot/blob/main/lerobot/motors/feetech/feetech.py) – for controlling Feetech servos
+- [`DynamixelMotorsBus`](https://github.com/huggingface/lerobot/blob/main/lerobot/motors/dynamixel/dynamixel.py) – for controlling Dynamixel servos
+
+Please refer to the [`MotorsBus`](https://github.com/huggingface/lerobot/blob/main/lerobot/motors/motors_bus.py) abstract class to learn about its API.
+For a good example of how it can be used, you can have a look at our own [SO101 follower implementation](https://github.com/huggingface/lerobot/blob/main/lerobot/robots/so101_follower/so101_follower.py)
+
+Use these if compatible. Otherwise, you'll need to find or write a Python interface (not covered in this tutorial):
+- Find an existing SDK in Python (or use bindings to C/C++)
+- Or implement a basic communication wrapper (e.g., via pyserial, socket, or CANopen)
+
+You're not alone—many community contributions use custom boards or firmware!
+
+For Feetech and Dynamixel, we currently support these servos:
+    - Feetech:
+        - STS & SMS series (protocol 0): `sts3215`, `sts3250`, `sm8512bl`
+        - SCS series (protocol 1): `scs0009`
+    - Dynamixel (protocol 2.0 only): `xl330-m077`, `xl330-m288`, `xl430-w250`, `xm430-w350`, `xm540-w270`, `xc430-w150`
+
+If you are using Feetech or Dynamixel servos that are not in this list, you can add those in the [Feetech table](https://github.com/huggingface/lerobot/blob/main/lerobot/motors/feetech/tables.py) or [Dynamixel table](https://github.com/huggingface/lerobot/blob/main/lerobot/motors/dynamixel/tables.py). Depending on the model, this will require you to add model-specific information. In most cases though, there shouldn't be a lot of additions to do.
+
+In the next sections, we'll use a `FeetechMotorsBus` as the motors interface for the examples. Replace it and adapt to your motors if necessary.
+
+## Step 1: Subclass the `Robot` Interface
+
+You’ll first need to specify the config class and a string identifier (`name`) for your robot. If your robot has special needs that you'd like to be able to change easily, it should go here (e.g. port/address, baudrate).
+
+Here, we'll add the port name and one camera by default for our robot:
+```python
+from dataclasses import dataclass, field
+
+from lerobot.cameras import CameraConfig
+from lerobot.cameras.opencv import OpenCVCameraConfig
+from lerobot.robots import RobotConfig
+
+
+@RobotConfig.register_subclass("my_cool_robot")
+@dataclass
+class MyCoolRobotConfig(RobotConfig):
+    port: str
+    cameras: dict[str, CameraConfig] = field(
+        default_factory={
+            "cam_1": OpenCVCameraConfig(
+                index_or_path=2,
+                fps=30,
+                width=480,
+                height=640,
+            ),
+        }
+    )
+```
+
+Have a look at our [Cameras tutorial](./cameras) to understand how to detect and add your camera.
+
+Next, we'll create our actual robot class which inherits from `Robot`. This abstract class defines a contract you must follow for your robot to be usable with the rest of the LeRobot tools.
+
+Here we'll create a simple 5-DoF robot with one camera. It could be a simple arm but notice that the `Robot` abstract class does not assume anything on your robot's form factor. You can let you imagination run wild when designing new robots!
+
+```python
+from lerobot.cameras import make_cameras_from_configs
+from lerobot.motors import Motor, MotorNormMode
+from lerobot.motors.feetech import FeetechMotorsBus
+from lerobot.robots import Robot
+
+class MyCoolRobot(Robot):
+    config_class = MyCoolRobotConfig
+    name = "my_cool_robot"
+
+    def __init__(self, config: MyCoolRobotConfig):
+        super().__init__(config)
+        self.bus = FeetechMotorsBus(
+            port=self.config.port,
+            motors={
+                "joint_1": Motor(1, "sts3250", MotorNormMode.RANGE_M100_100),
+                "joint_2": Motor(2, "sts3215", MotorNormMode.RANGE_M100_100),
+                "joint_3": Motor(3, "sts3215", MotorNormMode.RANGE_M100_100),
+                "joint_4": Motor(4, "sts3215", MotorNormMode.RANGE_M100_100),
+                "joint_5": Motor(5, "sts3215", MotorNormMode.RANGE_M100_100),
+            },
+            calibration=self.calibration,
+        )
+        self.cameras = make_cameras_from_configs(config.cameras)
+```
+
+## Step 2: Define Observation and Action Features
+
+These two properties define the *interface contract* between your robot and tools that consume it (such as data collection or learning pipelines).
+
+> [!WARNING]
+> Note that these properties must be callable even if the robot is not yet connected, so avoid relying on runtime hardware state to define them.
+
+### `observation_features`
+
+This property should return a dictionary describing the structure of sensor outputs from your robot. The keys match what `get_observation()` returns, and the values describe either the shape (for arrays/images) or the type (for simple values).
+
+Example for our 5-DoF arm with one camera:
+```python
+@property
+def _motors_ft(self) -> dict[str, type]:
+    return {
+        "joint_1.pos": float,
+        "joint_2.pos": float,
+        "joint_3.pos": float,
+        "joint_4.pos": float,
+        "joint_5.pos": float,
+    }
+
+@property
+def _cameras_ft(self) -> dict[str, tuple]:
+    return {
+        cam: (self.cameras[cam].height, self.cameras[cam].width, 3) for cam in self.cameras
+    }
+
+@property
+def observation_features(self) -> dict:
+    return {**self._motors_ft, **self._cameras_ft}
+```
+In this case, observations consist of a simple dict storing each motor's position and a camera image.
+
+### `action_features`
+
+This property describes the commands your robot expects via `send_action()`. Again, keys must match the expected input format, and values define the shape/type of each command.
+
+Here, we simply use the same joints proprioceptive features (`self._motors_ft`) as with `observation_features`: the action sent will simply the goal position for each motor.
+```python
+def action_features(self) -> dict:
+    return self._motors_ft
+```
+
+## Step 3: Handle Connection and Disconnection
+
+These methods should handle opening and closing communication with your hardware (e.g. serial ports, CAN interfaces, USB devices, cameras).
+
+### `is_connected`
+
+This property should simply reflect that communication with the robot's hardware is established. When this property is `True`, it should be possible to read and write to the hardware using `get_observation()` and `send_action()`.
+
+```python
+@property
+def is_connected(self) -> bool:
+    return self.bus.is_connected and all(cam.is_connected for cam in self.cameras.values())
+```
+
+### `connect()`
+
+This method should establish communication with the hardware. Moreover, if your robot needs calibration and is not calibrated, it should start a calibration procedure by default. If your robot needs some specific configuration, this should also be called here.
+
+```python
+def connect(self, calibrate: bool = True) -> None:
+    self.bus.connect()
+    if not self.is_calibrated and calibrate:
+        self.calibrate()
+
+    for cam in self.cameras.values():
+        cam.connect()
+
+    self.configure()
+```
+
+### `disconnect()`
+
+This method should gracefully terminate communication with the hardware: free any related resources (threads or processes), close ports, etc.
+
+Here, we already handle this in our `MotorsBus` and `Camera` classes so we just need to call their own `disconnect()` methods:
+```python
+def disconnect(self) -> None:
+    self.bus.disconnect()
+    for cam in self.cameras.values():
+        cam.disconnect()
+```
+
+## Step 4: Support Calibration and Configuration
+
+LeRobot supports saving and loading calibration data automatically. This is useful for joint offsets, zero positions, or sensor alignment.
+
+> Note that depending on your hardware, this may not apply. If that's the case, you can simply leave these methods as no-ops:
+> ```python
+> @property
+> def is_calibrated(self) -> bool:
+>    return True
+>
+> def calibrate(self) -> None:
+>    pass
+> ```
+
+### `is_calibrated`
+
+This should reflect whether your robot has the required calibration loaded.
+
+```python
+@property
+def is_calibrated(self) -> bool:
+    return self.bus.is_calibrated
+```
+
+### `calibrate()`
+
+The goal of the calibration is twofold:
+    - Know the physical range of motion of each motors in order to only send commands within this range.
+    - Normalize raw motors positions to sensible continuous values (e.g. percentages, degrees) instead of arbitrary discrete value dependant on the specific motor used that will not replicate elsewhere.
+
+It should implement the logic for calibration (if relevant) and update the `self.calibration` dictionary. If you are using Feetech or Dynamixel motors, our bus interfaces already include methods to help with this.
+
+```python
+def calibrate(self) -> None:
+    self.bus.disable_torque()
+    for motor in self.bus.motors:
+        self.bus.write("Operating_Mode", motor, OperatingMode.POSITION.value)
+
+    input(f"Move {self} to the middle of its range of motion and press ENTER....")
+    homing_offsets = self.bus.set_half_turn_homings()
+
+    print(
+        "Move all joints sequentially through their entire ranges "
+        "of motion.\nRecording positions. Press ENTER to stop..."
+    )
+    range_mins, range_maxes = self.bus.record_ranges_of_motion()
+
+    self.calibration = {}
+    for motor, m in self.bus.motors.items():
+        self.calibration[motor] = MotorCalibration(
+            id=m.id,
+            drive_mode=0,
+            homing_offset=homing_offsets[motor],
+            range_min=range_mins[motor],
+            range_max=range_maxes[motor],
+        )
+
+    self.bus.write_calibration(self.calibration)
+    self._save_calibration()
+    print("Calibration saved to", self.calibration_fpath)
+```
+
+### `configure()`
+
+Use this to set up any configuration for your hardware (servos control modes, controller gains, etc.). This should usually be run at connection time and be idempotent.
+
+```python
+def configure(self) -> None:
+    with self.bus.torque_disabled():
+        self.bus.configure_motors()
+        for motor in self.bus.motors:
+            self.bus.write("Operating_Mode", motor, OperatingMode.POSITION.value)
+            self.bus.write("P_Coefficient", motor, 16)
+            self.bus.write("I_Coefficient", motor, 0)
+            self.bus.write("D_Coefficient", motor, 32)
+```
+
+## Step 5: Implement Sensors Reading and Action Sending
+
+These are the most important runtime functions: the core I/O loop.
+
+### `get_observation()`
+
+Returns a dictionary of sensor values from the robot. These typically include motor states, camera frames, various sensors, etc. In the LeRobot framework, these observations are what will be fed to a policy in order to predict the actions to take. The dictionary keys and structure must match `observation_features`.
+
+```python
+def get_observation(self) -> dict[str, Any]:
+    if not self.is_connected:
+        raise ConnectionError(f"{self} is not connected.")
+
+    # Read arm position
+    obs_dict = self.bus.sync_read("Present_Position")
+    obs_dict = {f"{motor}.pos": val for motor, val in obs_dict.items()}
+
+    # Capture images from cameras
+    for cam_key, cam in self.cameras.items():
+        obs_dict[cam_key] = cam.async_read()
+
+    return obs_dict
+```
+
+### `send_action()`
+
+Takes a dictionary that matches `action_features`, and sends it to your hardware. You can add safety limits (clipping, smoothing) and return what was actually sent.
+
+For simplicity, we won't be adding any modification of the actions in our example here.
+
+```python
+def send_action(self, action: dict[str, Any]) -> dict[str, Any]:
+    goal_pos = {key.removesuffix(".pos"): val for key, val in action.items()}
+
+    # Send goal position to the arm
+    self.bus.sync_write("Goal_Position", goal_pos)
+
+    return action
+```
+
+## Adding a Teleoperator
+
+For implementing teleoperation devices, we also provide a [`Teleoperator`](https://github.com/huggingface/lerobot/blob/main/lerobot/teleoperators/teleoperator.py) base class. This class is very similar to the `Robot` base class and also doesn't assume anything on form factor.
+
+The main differences are in the I/O functions: a teleoperator allows you to produce action via `get_action` and can receive feedback actions via `send_feedback`. Feedback could be anything controllable on the teleoperation device that could help the person controlling it understand the consequences of the actions sent. Think motion/force feedback on a leader arm, vibrations on a gamepad controller for example. To implement a teleoperator, you can follow this same tutorial and adapt it for these two methods.
+
+## Wrapping Up
+
+Once your robot class is complete, you can leverage the LeRobot ecosystem:
+
+- Control your robot with available teleoperators or integrate directly your teleoperating device
+- Record training data and visualize it
+- Integrate it into RL or imitation learning pipelines
+
+Don't hesitate to reach out to the community for help on our [Discord](https://discord.gg/s3KuuzsPFb) 🤗

docs/source/koch.mdx

@@ -0,0 +1 @@
+../../src/lerobot/robots/koch_follower/koch.mdx

docs/source/lekiwi.mdx

@@ -0,0 +1 @@
+../../src/lerobot/robots/lekiwi/lekiwi.mdx

docs/source/notebooks.mdx

@@ -0,0 +1,29 @@
+# 🤗 LeRobot Notebooks
+
+This repository contains example notebooks for using LeRobot. These notebooks demonstrate how to train policies on real or simulation datasets using standardized policies.
+
+---
+
+### Training ACT
+
+[ACT](https://huggingface.co/papers/2304.13705) (Action Chunking Transformer) is a transformer-based policy architecture for imitation learning that processes robot states and camera inputs to generate smooth, chunked action sequences.
+
+We provide a ready-to-run Google Colab notebook to help you train ACT policies using datasets from the Hugging Face Hub, with optional logging to Weights & Biases.
+
+| Notebook | Colab |
+|:---------|:------|
+| [Train ACT with LeRobot](https://github.com/huggingface/notebooks/blob/main/lerobot/training-act.ipynb) | [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/lerobot/training-act.ipynb) |
+
+Expected training time for 100k steps: ~1.5 hours on an NVIDIA A100 GPU with batch size of `64`.
+
+### Training SmolVLA
+
+[SmolVLA](https://huggingface.co/papers/2506.01844) is a small but efficient Vision-Language-Action model. It is compact in size with 450 M-parameter and is developed by Hugging Face.
+
+We provide a ready-to-run Google Colab notebook to help you train SmolVLA policies using datasets from the Hugging Face Hub, with optional logging to Weights & Biases.
+
+| Notebook                                                                                                        | Colab                                                                                                                                                                                 |
+| :-------------------------------------------------------------------------------------------------------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| [Train SmolVLA with LeRobot](https://github.com/huggingface/notebooks/blob/main/lerobot/training-smolvla.ipynb) | [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/lerobot/training-smolvla.ipynb) |
+
+Expected training time for 20k steps: ~5 hours on an NVIDIA A100 GPU with batch size of `64`.

docs/source/smolvla.mdx

@@ -0,0 +1,97 @@
+# Finetune SmolVLA
+
+SmolVLA is Hugging Face’s lightweight foundation model for robotics. Designed for easy fine-tuning on LeRobot datasets, it helps accelerate your development!
+
+<p align="center">
+  <img src="https://cdn-uploads.huggingface.co/production/uploads/640e21ef3c82bd463ee5a76d/aooU0a3DMtYmy_1IWMaIM.png" alt="SmolVLA architecture." width="500"/>
+  <br/>
+  <em>Figure 1. SmolVLA takes as input (i) multiple cameras views, (ii) the robot’s current sensorimotor state, and (iii) a natural language instruction, encoded into contextual features used to condition the action expert when generating an action chunk.</em>
+</p>
+
+## Set Up Your Environment
+
+1. Install LeRobot by following our [Installation Guide](./installation).
+2. Install SmolVLA dependencies by running:
+
+   ```bash
+   pip install -e ".[smolvla]"
+   ```
+
+## Collect a dataset
+
+SmolVLA is a base model, so fine-tuning on your own data is required for optimal performance in your setup.
+We recommend recording ~50 episodes of your task as a starting point. Follow our guide to get started: [Recording a Dataset](https://huggingface.co/docs/lerobot/getting_started_real_world_robot#record-a-dataset)
+
+<Tip>
+
+In your dataset, make sure to have enough demonstrations per each variation (e.g. the cube position on the table if it is cube pick-place task) you are introducing.
+
+We recommend checking out the dataset linked below for reference that was used in the [SmolVLA paper](https://huggingface.co/papers/2506.01844):
+
+🔗 [SVLA SO100 PickPlace](https://huggingface.co/spaces/lerobot/visualize_dataset?path=%2Flerobot%2Fsvla_so100_pickplace%2Fepisode_0)
+
+In this dataset, we recorded 50 episodes across 5 distinct cube positions. For each position, we collected 10 episodes of pick-and-place interactions. This structure, repeating each variation several times, helped the model generalize better. We tried similar dataset with 25 episodes, and it was not enough leading to a bad performance. So, the data quality and quantity is definitely a key.
+After you have your dataset available on the Hub, you are good to go to use our finetuning script to adapt SmolVLA to your application.
+</Tip>
+
+## Finetune SmolVLA on your data
+
+Use [`smolvla_base`](https://hf.co/lerobot/smolvla_base), our pretrained 450M model, and fine-tune it on your data.
+Training the model for 20k steps will roughly take ~4 hrs on a single A100 GPU. You should tune the number of steps based on performance and your use-case.
+
+If you don't have a gpu device, you can train using our notebook on [![Google Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/notebooks/blob/main/lerobot/training-smolvla.ipynb)
+
+Pass your dataset to the training script using `--dataset.repo_id`. If you want to test your installation, run the following command where we use one of the datasets we collected for the [SmolVLA Paper](https://huggingface.co/papers/2506.01844).
+
+```bash
+cd lerobot && python -m lerobot.scripts.train \
+  --policy.path=lerobot/smolvla_base \
+  --dataset.repo_id=${HF_USER}/mydataset \
+  --batch_size=64 \
+  --steps=20000 \
+  --output_dir=outputs/train/my_smolvla \
+  --job_name=my_smolvla_training \
+  --policy.device=cuda \
+  --wandb.enable=true
+```
+
+<Tip>
+You can start with a small batch size and increase it incrementally, if the GPU allows it, as long as loading times remain short.
+</Tip>
+
+Fine-tuning is an art. For a complete overview of the options for finetuning, run
+
+```bash
+python -m lerobot.scripts.train --help
+```
+
+<p align="center">
+  <img src="https://cdn-uploads.huggingface.co/production/uploads/640e21ef3c82bd463ee5a76d/S-3vvVCulChREwHDkquoc.gif" alt="Comparison of SmolVLA across task variations." width="500"/>
+  <br/>
+  <em>Figure 2: Comparison of SmolVLA across task variations. From left to right: (1) pick-place cube counting, (2) pick-place cube counting, (3) pick-place cube counting under perturbations, and (4) generalization on pick-and-place of the lego block with real-world SO101.</em>
+</p>
+
+
+## Evaluate the finetuned model and run it in real-time
+
+Similarly for when recording an episode, it is recommended that you are logged in to the HuggingFace Hub. You can follow the corresponding steps: [Record a dataset](./getting_started_real_world_robot#record-a-dataset).
+Once you are logged in, you can run inference in your setup by doing:
+
+```bash
+python -m lerobot.record \
+  --robot.type=so101_follower \
+  --robot.port=/dev/ttyACM0 \ # <- Use your port
+  --robot.id=my_blue_follower_arm \ # <- Use your robot id
+  --robot.cameras="{ front: {type: opencv, index_or_path: 8, width: 640, height: 480, fps: 30}}" \ # <- Use your cameras
+  --dataset.single_task="Grasp a lego block and put it in the bin." \ # <- Use the same task description you used in your dataset recording
+  --dataset.repo_id=${HF_USER}/eval_DATASET_NAME_test \  # <- This will be the dataset name on HF Hub
+  --dataset.episode_time_s=50 \
+  --dataset.num_episodes=10 \
+  # <- Teleop optional if you want to teleoperate in between episodes \
+  # --teleop.type=so100_leader \
+  # --teleop.port=/dev/ttyACM0 \
+  # --teleop.id=my_red_leader_arm \
+  --policy.path=HF_USER/FINETUNE_MODEL_NAME # <- Use your fine-tuned model
+```
+
+Depending on your evaluation setup, you can configure the duration and the number of episodes to record for your evaluation suite.

docs/source/so100.mdx

@@ -0,0 +1 @@
+../../src/lerobot/robots/so100_follower/so100.mdx

docs/source/so101.mdx

@@ -0,0 +1 @@
+../../src/lerobot/robots/so101_follower/so101.mdx

examples/10_use_so100.md

@@ -1,622 +0,0 @@
-# Using the [SO-100](https://github.com/TheRobotStudio/SO-ARM100) with LeRobot
-
-## Table of Contents
-
-  - [A. Source the parts](#a-source-the-parts)
-  - [B. Install LeRobot](#b-install-lerobot)
-  - [C. Configure the Motors](#c-configure-the-motors)
-  - [D. Step-by-Step Assembly Instructions](#d-step-by-step-assembly-instructions)
-  - [E. Calibrate](#e-calibrate)
-  - [F. Teleoperate](#f-teleoperate)
-  - [G. Record a dataset](#g-record-a-dataset)
-  - [H. Visualize a dataset](#h-visualize-a-dataset)
-  - [I. Replay an episode](#i-replay-an-episode)
-  - [J. Train a policy](#j-train-a-policy)
-  - [K. Evaluate your policy](#k-evaluate-your-policy)
-  - [L. More Information](#l-more-information)
-
-## A. Source the parts
-
-Follow this [README](https://github.com/TheRobotStudio/SO-ARM100). It contains the bill of materials, with a link to source the parts, as well as the instructions to 3D print the parts,
-and advice if it's your first time printing or if you don't own a 3D printer.
-
-Before assembling, you will first need to configure your motors. To this end, we provide a nice script, so let's first install LeRobot. After configuration, we will also guide you through assembly.
-
-## B. Install LeRobot
-
-> [!TIP]
-> We use the Command Prompt (cmd) quite a lot. If you are not comfortable using the cmd or want to brush up using the command line you can have a look here: [Command line crash course](https://developer.mozilla.org/en-US/docs/Learn_web_development/Getting_started/Environment_setup/Command_line)
-
-On your computer:
-
-#### 1. [Install Miniconda](https://docs.anaconda.com/miniconda/install/#quick-command-line-install):
-
-#### 2. Restart shell
-Copy paste in your shell: `source ~/.bashrc` or for Mac: `source ~/.bash_profile` or `source ~/.zshrc` if you're using zshell
-
-#### 3. Create and activate a fresh conda environment for lerobot
-
-<details>
-<summary><strong>Video install instructions</strong></summary>
-
-<video src="https://github.com/user-attachments/assets/17172d3b-3b64-4b80-9cf1-b2b7c5cbd236"></video>
-
-</details>
-
-```bash
-conda create -y -n lerobot python=3.10
-```
-
-Then activate your conda environment (do this each time you open a shell to use lerobot!):
-```bash
-conda activate lerobot
-```
-
-#### 4. Clone LeRobot:
-```bash
-git clone https://github.com/huggingface/lerobot.git ~/lerobot
-```
-
-#### 5. Install ffmpeg in your environment:
-When using `miniconda`, install `ffmpeg` in your environment:
-```bash
-conda install ffmpeg -c conda-forge
-```
-
-#### 6. Install LeRobot with dependencies for the feetech motors:
-```bash
-cd ~/lerobot && pip install -e ".[feetech]"
-```
-
-Great :hugs:! You are now done installing LeRobot and we can begin assembling the SO100 arms :robot:.
-Every time you now want to use LeRobot you can go to the `~/lerobot` folder where we installed LeRobot and run one of the commands.
-
-## C. Configure the motors
-
-> [!NOTE]
-> Throughout this tutorial you will find videos on how to do the steps, the full video tutorial can be found here: [assembly video](https://www.youtube.com/watch?v=FioA2oeFZ5I).
-
-### 1. Find the USB ports associated to each arm
-
-Designate one bus servo adapter and 6 motors for your leader arm, and similarly the other bus servo adapter and 6 motors for the follower arm. It's convenient to label them and write on each motor if it's for the follower `F` or for the leader `L` and it's ID from 1 to 6 (F1...F6 and L1...L6).
-
-#### a. Run the script to find port
-
-<details>
-<summary><strong>Video finding port</strong></summary>
-  <video src="https://github.com/user-attachments/assets/4a21a14d-2046-4805-93c4-ee97a30ba33f"></video>
-  <video src="https://github.com/user-attachments/assets/1cc3aecf-c16d-4ff9-aec7-8c175afbbce2"></video>
-</details>
-
-To find the port for each bus servo adapter, run the utility script:
-```bash
-python lerobot/scripts/find_motors_bus_port.py
-```
-
-#### b. Example outputs
-
-Example output when identifying the leader arm's port (e.g., `/dev/tty.usbmodem575E0031751` on Mac, or possibly `/dev/ttyACM0` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this MotorsBus is /dev/tty.usbmodem575E0031751
-Reconnect the usb cable.
-```
-Example output when identifying the follower arm's port (e.g., `/dev/tty.usbmodem575E0032081`, or possibly `/dev/ttyACM1` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this MotorsBus is /dev/tty.usbmodem575E0032081
-Reconnect the usb cable.
-```
-
-#### c. Troubleshooting
-On Linux, you might need to give access to the USB ports by running:
-```bash
-sudo chmod 666 /dev/ttyACM0
-sudo chmod 666 /dev/ttyACM1
-```
-
-#### d. Update config file
-
-IMPORTANTLY: Now that you have your ports, update the **port** default values of [`SO100RobotConfig`](../lerobot/common/robot_devices/robots/configs.py). You will find something like:
-```python
-@RobotConfig.register_subclass("so100")
-@dataclass
-class So100RobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/so100"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-```
-
-### 2. Assembling the Base
-Let's begin with assembling the follower arm base
-
-#### a. Set IDs for all 12 motors
-
-<details>
-<summary><strong>Video configuring motor</strong></summary>
-  <video src="https://github.com/user-attachments/assets/ef9b3317-2e11-4858-b9d3-f0a02fb48ecf"></video>
-  <video src="https://github.com/user-attachments/assets/f36b5ed5-c803-4ebe-8947-b39278776a0d"></video>
-</details>
-
-Plug your first motor F1 and run this script to set its ID to 1. It will also set its present position to 2048, so expect your motor to rotate. Replace the text after --port to the corresponding follower control board port and run this command in cmd:
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 1
-```
-
-> [!NOTE]
-> These motors are currently limited. They can take values between 0 and 4096 only, which corresponds to a full turn. They can't turn more than that. 2048 is at the middle of this range, so we can take -2048 steps (180 degrees anticlockwise) and reach the maximum range, or take +2048 steps (180 degrees clockwise) and reach the maximum range. The configuration step also sets the homing offset to 0, so that if you misassembled the arm, you can always update the homing offset to account for a shift up to ± 2048 steps (± 180 degrees).
-
-Then unplug your motor and plug the second motor and set its ID to 2.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 2
-```
-
-Redo the process for all your motors until ID 6. Do the same for the 6 motors of the leader arm.
-
-
-#### b. Remove the gears of the 6 leader motors
-
-<details>
-<summary><strong>Video removing gears</strong></summary>
-
-<video src="https://github.com/user-attachments/assets/0c95b88c-5b85-413d-ba19-aee2f864f2a7"></video>
-
-</details>
-
-
-Follow the video for removing gears. You need to remove the gear for the motors of the leader arm. As a result, you will only use the position encoding of the motor and reduce friction to more easily operate the leader arm.
-
-## D. Step-by-Step Assembly Instructions
-
-**Step 1: Clean Parts**
-- Remove all support material from the 3D-printed parts.
----
-
-### Additional Guidance
-
-<details>
-<summary><strong>Video assembling arms</strong></summary>
-
-<video src="https://github.com/user-attachments/assets/488a39de-0189-4461-9de3-05b015f90cca"></video>
-
-</details>
-
-**Note:**
-This video provides visual guidance for assembling the arms, but it doesn't specify when or how to do the wiring. Inserting the cables beforehand is much easier than doing it afterward. The first arm may take a bit more than 1 hour to assemble, but once you get used to it, you can assemble the second arm in under 1 hour.
-
----
-
-### First Motor
-
-**Step 2: Insert Wires**
-- Insert two wires into the first motor.
-
-  <img src="../media/tutorial/img1.jpg" style="height:300px;">
-
-**Step 3: Install in Base**
-- Place the first motor into the base.
-
-  <img src="../media/tutorial/img2.jpg" style="height:300px;">
-
-**Step 4: Secure Motor**
-- Fasten the motor with 4 screws. Two from the bottom and two from top.
-
-**Step 5: Attach Motor Holder**
-- Slide over the first motor holder and fasten it using two screws (one on each side).
-
-  <img src="../media/tutorial/img4.jpg" style="height:300px;">
-
-**Step 6: Attach Motor Horns**
-- Install both motor horns, securing the top horn with a screw. Try not to move the motor position when attaching the motor horn, especially for the leader arms, where we removed the gears.
-
-  <img src="../media/tutorial/img5.jpg" style="height:300px;">
-<details>
-  <summary><strong>Video adding motor horn</strong></summary>
-  <video src="https://github.com/user-attachments/assets/ef3391a4-ad05-4100-b2bd-1699bf86c969"></video>
-</details>
-
-**Step 7: Attach Shoulder Part**
-- Route one wire to the back of the robot and the other to the left or in photo towards you (see photo).
-- Attach the shoulder part.
-
-  <img src="../media/tutorial/img6.jpg" style="height:300px;">
-
-**Step 8: Secure Shoulder**
-- Tighten the shoulder part with 4 screws on top and 4 on the bottom
-*(access bottom holes by turning the shoulder).*
-
----
-
-### Second Motor Assembly
-
-**Step 9: Install Motor 2**
-- Slide the second motor in from the top and link the wire from motor 1 to motor 2.
-
-  <img src="../media/tutorial/img8.jpg" style="height:300px;">
-
-**Step 10: Attach Shoulder Holder**
-- Add the shoulder motor holder.
-- Ensure the wire from motor 1 to motor 2 goes behind the holder while the other wire is routed upward (see photo).
-- This part can be tight to assemble, you can use a workbench like the image or a similar setup to push the part around the motor.
-
-  <div style="display: flex;">
-    <img src="../media/tutorial/img9.jpg" style="height:250px;">
-    <img src="../media/tutorial/img10.jpg" style="height:250px;">
-    <img src="../media/tutorial/img12.jpg" style="height:250px;">
-  </div>
-
-**Step 11: Secure Motor 2**
-- Fasten the second motor with 4 screws.
-
-**Step 12: Attach Motor Horn**
-- Attach both motor horns to motor 2, again use the horn screw.
-
-**Step 13: Attach Base**
-- Install the base attachment using 2 screws.
-
-  <img src="../media/tutorial/img11.jpg" style="height:300px;">
-
-**Step 14: Attach Upper Arm**
-- Attach the upper arm with 4 screws on each side.
-
-  <img src="../media/tutorial/img13.jpg" style="height:300px;">
-
----
-
-### Third Motor Assembly
-
-**Step 15: Install Motor 3**
-- Route the motor cable from motor 2 through the cable holder to motor 3, then secure motor 3 with 4 screws.
-
-**Step 16: Attach Motor Horn**
-- Attach both motor horns to motor 3 and secure one again with a horn screw.
-
-  <img src="../media/tutorial/img14.jpg" style="height:300px;">
-
-**Step 17: Attach Forearm**
-- Connect the forearm to motor 3 using 4 screws on each side.
-
-  <img src="../media/tutorial/img15.jpg" style="height:300px;">
-
----
-
-### Fourth Motor Assembly
-
-**Step 18: Install Motor 4**
-- Slide in motor 4, attach the cable from motor 3, and secure the cable in its holder with a screw.
-
-  <div style="display: flex;">
-    <img src="../media/tutorial/img16.jpg" style="height:300px;">
-    <img src="../media/tutorial/img19.jpg" style="height:300px;">
-  </div>
-
-**Step 19: Attach Motor Holder 4**
-- Install the fourth motor holder (a tight fit). Ensure one wire is routed upward and the wire from motor 3 is routed downward (see photo).
-
-  <img src="../media/tutorial/img17.jpg" style="height:300px;">
-
-**Step 20: Secure Motor 4 & Attach Horn**
-- Fasten motor 4 with 4 screws and attach its motor horns, use for one a horn screw.
-
-  <img src="../media/tutorial/img18.jpg" style="height:300px;">
-
----
-
-### Wrist Assembly
-
-**Step 21: Install Motor 5**
-- Insert motor 5 into the wrist holder and secure it with 2 front screws.
-
-  <img src="../media/tutorial/img20.jpg" style="height:300px;">
-
-**Step 22: Attach Wrist**
-- Connect the wire from motor 4 to motor 5. And already insert the other wire for the gripper.
-- Secure the wrist to motor 4 using 4 screws on both sides.
-
-  <img src="../media/tutorial/img22.jpg" style="height:300px;">
-
-**Step 23: Attach Wrist Horn**
-- Install only one motor horn on the wrist motor and secure it with a horn screw.
-
-  <img src="../media/tutorial/img23.jpg" style="height:300px;">
-
----
-
-### Follower Configuration
-
-**Step 24: Attach Gripper**
-- Attach the gripper to motor 5.
-
-  <img src="../media/tutorial/img24.jpg" style="height:300px;">
-
-**Step 25: Install Gripper Motor**
-- Insert the gripper motor, connect the motor wire from motor 5 to motor 6, and secure it with 3 screws on each side.
-
-  <img src="../media/tutorial/img25.jpg" style="height:300px;">
-
-**Step 26: Attach Gripper Horn & Claw**
-- Attach the motor horns and again use a horn screw.
-- Install the gripper claw and secure it with 4 screws on both sides.
-
-  <img src="../media/tutorial/img26.jpg" style="height:300px;">
-
-**Step 27: Mount Controller**
-- Attach the motor controller on the back.
-
-  <div style="display: flex;">
-    <img src="../media/tutorial/img27.jpg" style="height:300px;">
-    <img src="../media/tutorial/img28.jpg" style="height:300px;">
-  </div>
-
-*Assembly complete – proceed to Leader arm assembly.*
-
----
-
-### Leader Configuration
-
-For the leader configuration, perform **Steps 1–23**. Make sure that you removed the motor gears from the motors.
-
-**Step 24: Attach Leader Holder**
-- Mount the leader holder onto the wrist and secure it with a screw.
-
-  <img src="../media/tutorial/img29.jpg" style="height:300px;">
-
-**Step 25: Attach Handle**
-- Attach the handle to motor 5 using 4 screws.
-
-  <img src="../media/tutorial/img30.jpg" style="height:300px;">
-
-**Step 26: Install Gripper Motor**
-- Insert the gripper motor, secure it with 3 screws on each side, attach a motor horn using a horn screw, and connect the motor wire.
-
-  <img src="../media/tutorial/img31.jpg" style="height:300px;">
-
-**Step 27: Attach Trigger**
-- Attach the follower trigger with 4 screws.
-
-  <img src="../media/tutorial/img32.jpg" style="height:300px;">
-
-**Step 28: Mount Controller**
-- Attach the motor controller on the back.
-
-  <div style="display: flex;">
-    <img src="../media/tutorial/img27.jpg" style="height:300px;">
-    <img src="../media/tutorial/img28.jpg" style="height:300px;">
-  </div>
-
-*Assembly complete – proceed to calibration.*
-
-
-## E. Calibrate
-
-Next, you'll need to calibrate your SO-100 robot to ensure that the leader and follower arms have the same position values when they are in the same physical position.
-The calibration process is very important because it allows a neural network trained on one SO-100 robot to work on another.
-
-#### Manual calibration of follower arm
-
-You will need to move the follower arm to these positions sequentially, note that the rotated position is on the right side of the robot and you have to open the gripper fully.
-
-| 1. Middle position | 2. Zero position                                                                                                                                       | 3. Rotated position                                                                                                                                             | 4. Rest position                                                                                                                                       |
-| ------------ |------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="../media/so101/follower_middle.webp?raw=true" alt="SO-101 leader arm middle position" title="SO-101 leader arm middle position" style="width:100%;"> | <img src="../media/so101/follower_zero.webp?raw=true" alt="SO-101 leader arm zero position" title="SO-101 leader arm zero position" style="width:100%;"> | <img src="../media/so101/follower_rotated.webp?raw=true" alt="SO-101 leader arm rotated position" title="SO-101 leader arm rotated position" style="width:100%;"> | <img src="../media/so101/follower_rest.webp?raw=true" alt="SO-101 leader arm rest position" title="SO-101 leader arm rest position" style="width:100%;"> |
-
-Make sure both arms are connected and run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_follower"]'
-```
-
-#### Manual calibration of leader arm
-You will also need to move the leader arm to these positions sequentially:
-
-| 1. Middle position | 2. Zero position                                                                                                                                       | 3. Rotated position                                                                                                                                             | 4. Rest position                                                                                                                                       |
-| ------------ |------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="../media/so101/leader_middle.webp?raw=true" alt="SO-100 leader arm middle position" title="SO-100 leader arm middle position" style="width:100%;"> | <img src="../media/so101/leader_zero.webp?raw=true" alt="SO-100 leader arm zero position" title="SO-100 leader arm zero position" style="width:100%;"> | <img src="../media/so101/leader_rotated.webp?raw=true" alt="SO-100 leader arm rotated position" title="SO-100 leader arm rotated position" style="width:100%;"> | <img src="../media/so101/leader_rest.webp?raw=true" alt="SO-100 leader arm rest position" title="SO-100 leader arm rest position" style="width:100%;"> |
-
-Run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_leader"]'
-```
-
-## F. Teleoperate
-
-**Simple teleop**
-Then you are ready to teleoperate your robot! Run this simple script (it won't connect and display the cameras):
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --robot.cameras='{}' \
-  --control.type=teleoperate
-```
-
-
-#### a. Teleop with displaying cameras
-Follow [this guide to setup your cameras](https://github.com/huggingface/lerobot/blob/main/examples/7_get_started_with_real_robot.md#c-add-your-cameras-with-opencvcamera). Then you will be able to display the cameras on your computer while you are teleoperating by running the following code. This is useful to prepare your setup before recording your first dataset.
-
-> **NOTE:** To visualize the data, enable `--control.display_data=true`. This streams the data using `rerun`.
-
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --control.type=teleoperate
-```
-
-## G. Record a dataset
-
-Once you're familiar with teleoperation, you can record your first dataset with SO-100.
-
-If you want to use the Hugging Face hub features for uploading your dataset and you haven't previously done it, make sure you've logged in using a write-access token, which can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens):
-```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
-```
-
-Store your Hugging Face repository name in a variable to run these commands:
-```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
-echo $HF_USER
-```
-
-Record 2 episodes and upload your dataset to the hub:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/so100_test \
-  --control.tags='["so100","tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=2 \
-  --control.push_to_hub=true
-```
-
-Note: You can resume recording by adding `--control.resume=true`.
-
-## H. Visualize a dataset
-
-If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:
-```bash
-echo ${HF_USER}/so100_test
-```
-
-If you didn't upload with `--control.push_to_hub=false`, you can also visualize it locally with (a window can be opened in the browser `http://127.0.0.1:9090` with the visualization tool):
-```bash
-python lerobot/scripts/visualize_dataset_html.py \
-  --repo-id ${HF_USER}/so100_test \
-  --local-files-only 1
-```
-
-## I. Replay an episode
-
-Now try to replay the first episode on your robot:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --control.type=replay \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/so100_test \
-  --control.episode=0
-```
-
-## J. Train a policy
-
-To train a policy to control your robot, use the [`python lerobot/scripts/train.py`](../lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/so100_test \
-  --policy.type=act \
-  --output_dir=outputs/train/act_so100_test \
-  --job_name=act_so100_test \
-  --policy.device=cuda \
-  --wandb.enable=true
-```
-
-Let's explain it:
-1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/so100_test`.
-2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../lerobot/common/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
-4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
-5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
-
-Training should take several hours. You will find checkpoints in `outputs/train/act_so100_test/checkpoints`.
-
-To resume training from a checkpoint, below is an example command to resume from `last` checkpoint of the `act_so100_test` policy:
-```bash
-python lerobot/scripts/train.py \
-  --config_path=outputs/train/act_so100_test/checkpoints/last/pretrained_model/train_config.json \
-  --resume=true
-```
-
-## K. Evaluate your policy
-
-You can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) but with a policy checkpoint as input. For instance, run this command to record 10 evaluation episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so100 \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/eval_act_so100_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=10 \
-  --control.push_to_hub=true \
-  --control.policy.path=outputs/train/act_so100_test/checkpoints/last/pretrained_model
-```
-
-As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
-1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_act_so100_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_so100_test`).
-2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_so100_test`).
-
-## L. More Information
-
-Follow this [previous tutorial](https://github.com/huggingface/lerobot/blob/main/examples/7_get_started_with_real_robot.md#4-train-a-policy-on-your-data) for a more in-depth tutorial on controlling real robots with LeRobot.
-
-> [!TIP]
->  If you have any questions or need help, please reach out on [Discord](https://discord.com/invite/s3KuuzsPFb) in the channel [`#so100-arm`](https://discord.com/channels/1216765309076115607/1237741463832363039).

examples/11_use_lekiwi.md

@@ -1,597 +0,0 @@
-# Using the [LeKiwi](https://github.com/SIGRobotics-UIUC/LeKiwi) Robot with LeRobot
-
-## Table of Contents
-
-  - [A. Source the parts](#a-source-the-parts)
-  - [B. Install software Pi](#b-install-software-on-pi)
-  - [C. Setup LeRobot laptop/pc](#c-install-lerobot-on-laptop)
-  - [D. Assemble the arms](#d-assembly)
-  - [E. Calibrate](#e-calibration)
-  - [F. Teleoperate](#f-teleoperate)
-  - [G. Record a dataset](#g-record-a-dataset)
-  - [H. Visualize a dataset](#h-visualize-a-dataset)
-  - [I. Replay an episode](#i-replay-an-episode)
-  - [J. Train a policy](#j-train-a-policy)
-  - [K. Evaluate your policy](#k-evaluate-your-policy)
-
-> [!TIP]
->  If you have any questions or need help, please reach out on [Discord](https://discord.com/invite/s3KuuzsPFb) in the channel [`#mobile-so-100-arm`](https://discord.com/channels/1216765309076115607/1318390825528332371).
-
-## A. Source the parts
-
-Follow this [README](https://github.com/SIGRobotics-UIUC/LeKiwi). It contains the bill of materials, with a link to source the parts, as well as the instructions to 3D print the parts, and advice if it's your first time printing or if you don't own a 3D printer.
-
-Before assembling, you will first need to configure your motors. To this end, we provide a nice script, so let's first install LeRobot. After configuration, we will also guide you through assembly.
-
-### Wired version
-If you have the **wired** LeKiwi version you can skip the installation of the Raspberry Pi and setting up SSH. You can also run all commands directly on your PC for both the LeKiwi scripts and the leader arm scripts for teleoperating.
-
-## B. Install software on Pi
-Now we have to setup the remote PC that will run on the LeKiwi Robot. This is normally a Raspberry Pi, but can be any PC that can run on 5V and has enough usb ports (2 or more) for the cameras and motor control board.
-
-### Install OS
-For setting up the Raspberry Pi and its SD-card see: [Setup PI](https://www.raspberrypi.com/documentation/computers/getting-started.html). Here is explained how to download the [Imager](https://www.raspberrypi.com/software/) to install Raspberry Pi OS or Ubuntu.
-
-### Setup SSH
-After setting up your Pi, you should enable and setup [SSH](https://www.raspberrypi.com/news/coding-on-raspberry-pi-remotely-with-visual-studio-code/) (Secure Shell Protocol) so you can login into the Pi from your laptop without requiring a screen, keyboard and mouse in the Pi. A great tutorial on how to do this can be found [here](https://www.raspberrypi.com/documentation/computers/remote-access.html#ssh). Logging into your Pi can be done in your Command Prompt (cmd) or if you use VSCode you can use [this](https://marketplace.visualstudio.com/items?itemName=ms-vscode-remote.remote-ssh) extension.
-
-### Install LeRobot
-
-On your Raspberry Pi:
-
-#### 1. [Install Miniconda](https://docs.anaconda.com/miniconda/install/#quick-command-line-install):
-
-#### 2. Restart shell
-Copy paste in your shell: `source ~/.bashrc` or for Mac: `source ~/.bash_profile` or `source ~/.zshrc` if you're using zshell
-
-#### 3. Create and activate a fresh conda environment for lerobot
-
-<details>
-<summary><strong>Video install instructions</strong></summary>
-
-<video src="https://github.com/user-attachments/assets/17172d3b-3b64-4b80-9cf1-b2b7c5cbd236"></video>
-
-</details>
-
-```bash
-conda create -y -n lerobot python=3.10
-```
-
-Then activate your conda environment (do this each time you open a shell to use lerobot!):
-```bash
-conda activate lerobot
-```
-
-#### 4. Clone LeRobot:
-```bash
-git clone https://github.com/huggingface/lerobot.git ~/lerobot
-```
-
-#### 5. Install ffmpeg in your environment:
-When using `miniconda`, install `ffmpeg` in your environment:
-```bash
-conda install ffmpeg -c conda-forge
-```
-
-#### 6. Install LeRobot with dependencies for the feetech motors:
-```bash
-cd ~/lerobot && pip install -e ".[feetech]"
-```
-
-## C. Install LeRobot on laptop
-If you already have install LeRobot on your laptop you can skip this step, otherwise please follow along as we do the same steps we did on the Pi.
-
-> [!TIP]
-> We use the Command Prompt (cmd) quite a lot. If you are not comfortable using the cmd or want to brush up using the command line you can have a look here: [Command line crash course](https://developer.mozilla.org/en-US/docs/Learn_web_development/Getting_started/Environment_setup/Command_line)
-
-On your computer:
-
-#### 1. [Install Miniconda](https://docs.anaconda.com/miniconda/install/#quick-command-line-install):
-
-#### 2. Restart shell
-Copy paste in your shell: `source ~/.bashrc` or for Mac: `source ~/.bash_profile` or `source ~/.zshrc` if you're using zshell
-
-#### 3. Create and activate a fresh conda environment for lerobot
-
-<details>
-<summary><strong>Video install instructions</strong></summary>
-
-<video src="https://github.com/user-attachments/assets/17172d3b-3b64-4b80-9cf1-b2b7c5cbd236"></video>
-
-</details>
-
-```bash
-conda create -y -n lerobot python=3.10
-```
-
-Then activate your conda environment (do this each time you open a shell to use lerobot!):
-```bash
-conda activate lerobot
-```
-
-#### 4. Clone LeRobot:
-```bash
-git clone https://github.com/huggingface/lerobot.git ~/lerobot
-```
-
-#### 5. Install ffmpeg in your environment:
-When using `miniconda`, install `ffmpeg` in your environment:
-```bash
-conda install ffmpeg -c conda-forge
-```
-
-#### 6. Install LeRobot with dependencies for the feetech motors:
-```bash
-cd ~/lerobot && pip install -e ".[feetech]"
-```
-
-Great :hugs:! You are now done installing LeRobot and we can begin assembling the SO100 arms and Mobile base :robot:.
-Every time you now want to use LeRobot you can go to the `~/lerobot` folder where we installed LeRobot and run one of the commands.
-
-# D. Assembly
-
-First we will assemble the two SO100 arms. One to attach to the mobile base and one for teleoperation. Then we will assemble the mobile base.
-
-## SO100 Arms
-### Configure motors
-The instructions for configuring the motors can be found [Here](https://github.com/huggingface/lerobot/blob/main/examples/10_use_so100.md#c-configure-the-motors) in step C of the SO100 tutorial. Besides the ID's for the arm motors we also need to set the motor ID's for the mobile base. These need to be in a specific order to work. Below an image of the motor ID's and motor mounting positions for the mobile base. Note that we only use one Motor Control board on LeKiwi. This means the motor ID's for the wheels are 7, 8 and 9.
-
-<img src="../media/lekiwi/motor_ids.webp?raw=true" alt="Motor ID's for mobile robot" title="Motor ID's for mobile robot" width="60%">
-
-### Assemble arms
-[Assemble arms instruction](https://github.com/huggingface/lerobot/blob/main/examples/10_use_so100.md#d-assemble-the-arms)
-
-## Mobile base (LeKiwi)
-[Assemble LeKiwi](https://github.com/SIGRobotics-UIUC/LeKiwi)
-
-### Update config
-Both config files on the LeKiwi LeRobot and on the laptop should be the same. First we should find the Ip address of the Raspberry Pi of the mobile manipulator. This is the same Ip address used in SSH. We also need the usb port of the control board of the leader arm on the laptop and the port of the control board on LeKiwi. We can find these ports with the following script.
-
-#### a. Run the script to find port
-
-<details>
-<summary><strong>Video finding port</strong></summary>
-  <video src="https://github.com/user-attachments/assets/4a21a14d-2046-4805-93c4-ee97a30ba33f"></video>
-  <video src="https://github.com/user-attachments/assets/1cc3aecf-c16d-4ff9-aec7-8c175afbbce2"></video>
-</details>
-
-To find the port for each bus servo adapter, run the utility script:
-```bash
-python lerobot/scripts/find_motors_bus_port.py
-```
-
-#### b. Example outputs
-
-Example output when identifying the leader arm's port (e.g., `/dev/tty.usbmodem575E0031751` on Mac, or possibly `/dev/ttyACM0` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0031751
-Reconnect the usb cable.
-```
-Example output when identifying the follower arm's port (e.g., `/dev/tty.usbmodem575E0032081`, or possibly `/dev/ttyACM1` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0032081
-Reconnect the usb cable.
-```
-
-#### c. Troubleshooting
-On Linux, you might need to give access to the USB ports by running:
-```bash
-sudo chmod 666 /dev/ttyACM0
-sudo chmod 666 /dev/ttyACM1
-```
-
-#### d. Update config file
-
-IMPORTANTLY: Now that you have your ports of leader and follower arm and ip address of the mobile-so100, update the **ip** in Network configuration, **port** in leader_arms and **port** in lekiwi. In the [`LeKiwiRobotConfig`](../lerobot/common/robot_devices/robots/configs.py) file. Where you will find something like:
-```python
-@RobotConfig.register_subclass("lekiwi")
-@dataclass
-class LeKiwiRobotConfig(RobotConfig):
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    # Network Configuration
-    ip: str = "172.17.133.91"
-    port: int = 5555
-    video_port: int = 5556
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "mobile": OpenCVCameraConfig(camera_index="/dev/video0", fps=30, width=640, height=480),
-            "mobile2": OpenCVCameraConfig(camera_index="/dev/video2", fps=30, width=640, height=480),
-        }
-    )
-
-    calibration_dir: str = ".cache/calibration/lekiwi"
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0077581",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/ttyACM0",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                    "left_wheel": (7, "sts3215"),
-                    "back_wheel": (8, "sts3215"),
-                    "right_wheel": (9, "sts3215"),
-                },
-            ),
-        }
-    )
-
-    teleop_keys: dict[str, str] = field(
-        default_factory=lambda: {
-            # Movement
-            "forward": "w",
-            "backward": "s",
-            "left": "a",
-            "right": "d",
-            "rotate_left": "z",
-            "rotate_right": "x",
-            # Speed control
-            "speed_up": "r",
-            "speed_down": "f",
-            # quit teleop
-            "quit": "q",
-        }
-    )
-
-    mock: bool = False
-```
-
-## Wired version
-
-For the wired LeKiwi version your configured IP address should refer to your own laptop (127.0.0.1), because leader arm and LeKiwi are in this case connected to own laptop. Below and example configuration for this wired setup:
-```python
-@RobotConfig.register_subclass("lekiwi")
-@dataclass
-class LeKiwiRobotConfig(RobotConfig):
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    # Network Configuration
-    ip: str = "127.0.0.1"
-    port: int = 5555
-    video_port: int = 5556
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "front": OpenCVCameraConfig(
-                camera_index=0, fps=30, width=640, height=480, rotation=90
-            ),
-            "wrist": OpenCVCameraConfig(
-                camera_index=1, fps=30, width=640, height=480, rotation=180
-            ),
-        }
-    )
-
-    calibration_dir: str = ".cache/calibration/lekiwi"
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0077581",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431061",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                    "left_wheel": (7, "sts3215"),
-                    "back_wheel": (8, "sts3215"),
-                    "right_wheel": (9, "sts3215"),
-                },
-            ),
-        }
-    )
-
-    teleop_keys: dict[str, str] = field(
-        default_factory=lambda: {
-            # Movement
-            "forward": "w",
-            "backward": "s",
-            "left": "a",
-            "right": "d",
-            "rotate_left": "z",
-            "rotate_right": "x",
-            # Speed control
-            "speed_up": "r",
-            "speed_down": "f",
-            # quit teleop
-            "quit": "q",
-        }
-    )
-
-    mock: bool = False
-```
-
-# E. Calibration
-Now we have to calibrate the leader arm and the follower arm. The wheel motors don't have to be calibrated.
-
-
-### Calibrate follower arm (on mobile base)
-> [!IMPORTANT]
-> Contrarily to step 6 of the [assembly video](https://youtu.be/FioA2oeFZ5I?t=724) which illustrates the auto calibration, we will actually do manual calibration of follower for now.
-
-You will need to move the follower arm to these positions sequentially:
-
-| 1. Zero position                                                                                                                                                  | 2. Rotated position                                                                                                                                                        | 3. Rest position                                                                                                                                                  |
-| ----------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| <img src="../media/lekiwi/mobile_calib_zero.webp?raw=true" alt="SO-100 follower arm zero position" title="SO-100 follower arm zero position" style="width:100%;"> | <img src="../media/lekiwi/mobile_calib_rotated.webp?raw=true" alt="SO-100 follower arm rotated position" title="SO-100 follower arm rotated position" style="width:100%;"> | <img src="../media/lekiwi/mobile_calib_rest.webp?raw=true" alt="SO-100 follower arm rest position" title="SO-100 follower arm rest position" style="width:100%;"> |
-
-Make sure the arm is connected to the Raspberry Pi and run this script (on the Raspberry Pi) to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_follower"]'
-```
-
-### Wired version
-If you have the **wired** LeKiwi version please run all commands including this calibration command on your laptop.
-
-### Calibrate leader arm
-Then to calibrate the leader arm (which is attached to the laptop/pc). You will need to move the leader arm to these positions sequentially:
-
-| 1. Zero position                                                                                                                                       | 2. Rotated position                                                                                                                                             | 3. Rest position                                                                                                                                       |
-| ------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="../media/so100/leader_zero.webp?raw=true" alt="SO-100 leader arm zero position" title="SO-100 leader arm zero position" style="width:100%;"> | <img src="../media/so100/leader_rotated.webp?raw=true" alt="SO-100 leader arm rotated position" title="SO-100 leader arm rotated position" style="width:100%;"> | <img src="../media/so100/leader_rest.webp?raw=true" alt="SO-100 leader arm rest position" title="SO-100 leader arm rest position" style="width:100%;"> |
-
-Run this script (on your laptop/pc) to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_leader"]'
-```
-
-# F. Teleoperate
-
-> [!TIP]
-> If you're using a Mac, you might need to give Terminal permission to access your keyboard. Go to System Preferences > Security & Privacy > Input Monitoring and check the box for Terminal.
-
-To teleoperate SSH into your Raspberry Pi, and run `conda activate lerobot` and this script:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=remote_robot
-```
-
-Then on your laptop, also run `conda activate lerobot` and this script:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=teleoperate \
-  --control.fps=30
-```
-
-> **NOTE:** To visualize the data, enable `--control.display_data=true`. This streams the data using `rerun`. For the `--control.type=remote_robot` you will also need to set `--control.viewer_ip` and `--control.viewer_port`
-
-You should see on your laptop something like this: ```[INFO] Connected to remote robot at tcp://172.17.133.91:5555 and video stream at tcp://172.17.133.91:5556.``` Now you can move the leader arm and use the keyboard (w,a,s,d) to drive forward, left, backwards, right. And use (z,x) to turn left or turn right. You can use (r,f) to increase and decrease the speed of the mobile robot. There are three speed modes, see the table below:
-| Speed Mode | Linear Speed (m/s) | Rotation Speed (deg/s) |
-| ---------- | ------------------ | ---------------------- |
-| Fast       | 0.4                | 90                     |
-| Medium     | 0.25               | 60                     |
-| Slow       | 0.1                | 30                     |
-
-
-| Key | Action         |
-| --- | -------------- |
-| W   | Move forward   |
-| A   | Move left      |
-| S   | Move backward  |
-| D   | Move right     |
-| Z   | Turn left      |
-| X   | Turn right     |
-| R   | Increase speed |
-| F   | Decrease speed |
-
-> [!TIP]
->  If you use a different keyboard you can change the keys for each command in the [`LeKiwiRobotConfig`](../lerobot/common/robot_devices/robots/configs.py).
-
-### Wired version
-If you have the **wired** LeKiwi version please run all commands including both these teleoperation commands on your laptop.
-
-## Troubleshoot communication
-
-If you are having trouble connecting to the Mobile SO100, follow these steps to diagnose and resolve the issue.
-
-### 1. Verify IP Address Configuration
-Make sure that the correct ip for the Pi is set in the configuration file. To check the Raspberry Pi's IP address, run (on the Pi command line):
-```bash
-hostname -I
-```
-
-### 2. Check if Pi is reachable from laptop/pc
-Try pinging the Raspberry Pi from your laptop:
-```bach
-ping <your_pi_ip_address>
-```
-
-If the ping fails:
-- Ensure the Pi is powered on and connected to the same network.
-- Check if SSH is enabled on the Pi.
-
-### 3. Try SSH connection
-If you can't SSH into the Pi, it might not be properly connected. Use:
-```bash
-ssh <your_pi_user_name>@<your_pi_ip_address>
-```
-If you get a connection error:
-- Ensure SSH is enabled on the Pi by running:
-  ```bash
-  sudo raspi-config
-  ```
-  Then navigate to: **Interfacing Options -> SSH** and enable it.
-
-### 4. Same config file
-Make sure the configuration file on both your laptop/pc and the Raspberry Pi is the same.
-
-# G. Record a dataset
-Once you're familiar with teleoperation, you can record your first dataset with LeKiwi.
-
-To start the program on LeKiwi, SSH into your Raspberry Pi, and run `conda activate lerobot` and this script:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=remote_robot
-```
-
-If you want to use the Hugging Face hub features for uploading your dataset and you haven't previously done it, make sure you've logged in using a write-access token, which can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens):
-```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
-```
-
-Store your Hugging Face repository name in a variable to run these commands:
-```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
-echo $HF_USER
-```
-On your laptop then run this command to record 2 episodes and upload your dataset to the hub:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/lekiwi_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=2 \
-  --control.push_to_hub=true
-```
-
-Note: You can resume recording by adding `--control.resume=true`.
-
-### Wired version
-If you have the **wired** LeKiwi version please run all commands including both these record dataset commands on your laptop.
-
-# H. Visualize a dataset
-
-If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:
-```bash
-echo ${HF_USER}/lekiwi_test
-```
-
-If you didn't upload with `--control.push_to_hub=false`, you can also visualize it locally with (a window can be opened in the browser `http://127.0.0.1:9090` with the visualization tool):
-```bash
-python lerobot/scripts/visualize_dataset_html.py \
-  --repo-id ${HF_USER}/lekiwi_test \
-  --local-files-only 1
-```
-
-# I. Replay an episode
-Now try to replay the first episode on your robot:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=replay \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/lekiwi_test \
-  --control.episode=0
-```
-
-## J. Train a policy
-
-To train a policy to control your robot, use the [`python lerobot/scripts/train.py`](../lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/lekiwi_test \
-  --policy.type=act \
-  --output_dir=outputs/train/act_lekiwi_test \
-  --job_name=act_lekiwi_test \
-  --policy.device=cuda \
-  --wandb.enable=true
-```
-
-Let's explain it:
-1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/lekiwi_test`.
-2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../lerobot/common/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
-4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
-5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
-
-Training should take several hours. You will find checkpoints in `outputs/train/act_lekiwi_test/checkpoints`.
-
-## K. Evaluate your policy
-
-You can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) but with a policy checkpoint as input. For instance, run this command to record 10 evaluation episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Drive to the red block and pick it up" \
-  --control.repo_id=${HF_USER}/eval_act_lekiwi_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=10 \
-  --control.push_to_hub=true \
-  --control.policy.path=outputs/train/act_lekiwi_test/checkpoints/last/pretrained_model
-```
-
-As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
-1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_act_lekiwi_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_lekiwi_test`).
-2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_lekiwi_test`).

examples/11_use_moss.md

@@ -1,337 +0,0 @@
-This tutorial explains how to use [Moss v1](https://github.com/jess-moss/moss-robot-arms) with LeRobot.
-
-## Source the parts
-
-Follow this [README](https://github.com/jess-moss/moss-robot-arms). It contains the bill of materials with link to source the parts, as well as the instructions to 3D print the parts and advice if it's your first time printing or if you don't own a 3D printer already.
-
-**Important**: Before assembling, you will first need to configure your motors. To this end, we provide a nice script, so let's first install LeRobot. After configuration, we will also guide you through assembly.
-
-## Install LeRobot
-
-On your computer:
-
-1. [Install Miniconda](https://docs.anaconda.com/miniconda/#quick-command-line-install):
-```bash
-mkdir -p ~/miniconda3
-wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
-bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
-rm ~/miniconda3/miniconda.sh
-~/miniconda3/bin/conda init bash
-```
-
-2. Restart shell or `source ~/.bashrc`
-
-3. Create and activate a fresh conda environment for lerobot
-```bash
-conda create -y -n lerobot python=3.10 && conda activate lerobot
-```
-
-4. Clone LeRobot:
-```bash
-git clone https://github.com/huggingface/lerobot.git ~/lerobot
-```
-
-5. Install ffmpeg in your environment:
-When using `miniconda`, install `ffmpeg` in your environment:
-```bash
-conda install ffmpeg -c conda-forge
-```
-
-6. Install LeRobot with dependencies for the feetech motors:
-```bash
-cd ~/lerobot && pip install -e ".[feetech]"
-```
-
-## Configure the motors
-
-Follow step 1 of the [assembly video](https://www.youtube.com/watch?v=DA91NJOtMic) which illustrates the use of our scripts below.
-
-**Find USB ports associated to your arms**
-To find the correct ports for each arm, run the utility script twice:
-```bash
-python lerobot/scripts/find_motors_bus_port.py
-```
-
-Example output when identifying the leader arm's port (e.g., `/dev/tty.usbmodem575E0031751` on Mac, or possibly `/dev/ttyACM0` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0031751
-Reconnect the usb cable.
-```
-
-Example output when identifying the follower arm's port (e.g., `/dev/tty.usbmodem575E0032081`, or possibly `/dev/ttyACM1` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0032081
-Reconnect the usb cable.
-```
-
-Troubleshooting: On Linux, you might need to give access to the USB ports by running:
-```bash
-sudo chmod 666 /dev/ttyACM0
-sudo chmod 666 /dev/ttyACM1
-```
-
-#### Update config file
-
-IMPORTANTLY: Now that you have your ports, update the **port** default values of [`MossRobotConfig`](../lerobot/common/robot_devices/robots/configs.py). You will find something like:
-```python
-@RobotConfig.register_subclass("moss")
-@dataclass
-class MossRobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/moss"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-```
-
-**Configure your motors**
-Plug your first motor and run this script to set its ID to 1. It will also set its present position to 2048, so expect your motor to rotate:
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 1
-```
-
-Note: These motors are currently limitated. They can take values between 0 and 4096 only, which corresponds to a full turn. They can't turn more than that. 2048 is at the middle of this range, so we can take -2048 steps (180 degrees anticlockwise) and reach the maximum range, or take +2048 steps (180 degrees clockwise) and reach the maximum range. The configuration step also sets the homing offset to 0, so that if you misassembled the arm, you can always update the homing offset to account for a shift up to ± 2048 steps (± 180 degrees).
-
-Then unplug your motor and plug the second motor and set its ID to 2.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 2
-```
-
-Redo the process for all your motors until ID 6. Do the same for the 6 motors of the leader arm.
-
-**Remove the gears of the 6 leader motors**
-Follow step 2 of the [assembly video](https://www.youtube.com/watch?v=DA91NJOtMic). You need to remove the gear for the motors of the leader arm. As a result, you will only use the position encoding of the motor and reduce friction to more easily operate the leader arm.
-
-**Add motor horn to the motors**
-Follow step 3 of the [assembly video](https://www.youtube.com/watch?v=DA91NJOtMic). For Moss v1, you need to align the holes on the motor horn to the motor spline to be approximately 3, 6, 9 and 12 o'clock.
-Try to avoid rotating the motor while doing so to keep position 2048 set during configuration. It is especially tricky for the leader motors as it is more sensible without the gears, but it's ok if it's a bit rotated.
-
-## Assemble the arms
-
-Follow step 4 of the [assembly video](https://www.youtube.com/watch?v=DA91NJOtMic). The first arm should take a bit more than 1 hour to assemble, but once you get used to it, you can do it under 1 hour for the second arm.
-
-## Calibrate
-
-Next, you'll need to calibrate your Moss v1 robot to ensure that the leader and follower arms have the same position values when they are in the same physical position. This calibration is essential because it allows a neural network trained on one Moss v1 robot to work on another.
-
-**Manual calibration of follower arm**
-/!\ Contrarily to step 6 of the [assembly video](https://www.youtube.com/watch?v=DA91NJOtMic) which illustrates the auto calibration, we will actually do manual calibration of follower for now.
-
-You will need to move the follower arm to these positions sequentially:
-
-| 1. Zero position                                                                                                                                              | 2. Rotated position                                                                                                                                                    | 3. Rest position                                                                                                                                              |
-| ------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| <img src="../media/moss/follower_zero.webp?raw=true" alt="Moss v1 follower arm zero position" title="Moss v1 follower arm zero position" style="width:100%;"> | <img src="../media/moss/follower_rotated.webp?raw=true" alt="Moss v1 follower arm rotated position" title="Moss v1 follower arm rotated position" style="width:100%;"> | <img src="../media/moss/follower_rest.webp?raw=true" alt="Moss v1 follower arm rest position" title="Moss v1 follower arm rest position" style="width:100%;"> |
-
-Make sure both arms are connected and run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_follower"]'
-```
-
-**Manual calibration of leader arm**
-Follow step 6 of the [assembly video](https://www.youtube.com/watch?v=DA91NJOtMic) which illustrates the manual calibration. You will need to move the leader arm to these positions sequentially:
-
-| 1. Zero position                                                                                                                                        | 2. Rotated position                                                                                                                                              | 3. Rest position                                                                                                                                        |
-| ------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| <img src="../media/moss/leader_zero.webp?raw=true" alt="Moss v1 leader arm zero position" title="Moss v1 leader arm zero position" style="width:100%;"> | <img src="../media/moss/leader_rotated.webp?raw=true" alt="Moss v1 leader arm rotated position" title="Moss v1 leader arm rotated position" style="width:100%;"> | <img src="../media/moss/leader_rest.webp?raw=true" alt="Moss v1 leader arm rest position" title="Moss v1 leader arm rest position" style="width:100%;"> |
-
-Run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_leader"]'
-```
-
-## Teleoperate
-
-**Simple teleop**
-Then you are ready to teleoperate your robot! Run this simple script (it won't connect and display the cameras):
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --robot.cameras='{}' \
-  --control.type=teleoperate
-```
-
-
-**Teleop with displaying cameras**
-Follow [this guide to setup your cameras](https://github.com/huggingface/lerobot/blob/main/examples/7_get_started_with_real_robot.md#c-add-your-cameras-with-opencvcamera). Then you will be able to display the cameras on your computer while you are teleoperating by running the following code. This is useful to prepare your setup before recording your first dataset.
-
-> **NOTE:** To visualize the data, enable `--control.display_data=true`. This streams the data using `rerun`.
-
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --control.type=teleoperate
-```
-
-## Record a dataset
-
-Once you're familiar with teleoperation, you can record your first dataset with Moss v1.
-
-If you want to use the Hugging Face hub features for uploading your dataset and you haven't previously done it, make sure you've logged in using a write-access token, which can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens):
-```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
-```
-
-Store your Hugging Face repository name in a variable to run these commands:
-```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
-echo $HF_USER
-```
-
-Record 2 episodes and upload your dataset to the hub:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/moss_test \
-  --control.tags='["moss","tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=2 \
-  --control.push_to_hub=true
-```
-
-Note: You can resume recording by adding `--control.resume=true`.
-
-## Visualize a dataset
-
-If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:
-```bash
-echo ${HF_USER}/moss_test
-```
-
-If you didn't upload with `--control.push_to_hub=false`, you can also visualize it locally with:
-```bash
-python lerobot/scripts/visualize_dataset_html.py \
-  --repo-id ${HF_USER}/moss_test \
-  --local-files-only 1
-```
-
-## Replay an episode
-
-Now try to replay the first episode on your robot:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --control.type=replay \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/moss_test \
-  --control.episode=0
-```
-
-## Train a policy
-
-To train a policy to control your robot, use the [`python lerobot/scripts/train.py`](../lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/moss_test \
-  --policy.type=act \
-  --output_dir=outputs/train/act_moss_test \
-  --job_name=act_moss_test \
-  --policy.device=cuda \
-  --wandb.enable=true
-```
-
-Let's explain it:
-1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/moss_test`.
-2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../lerobot/common/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
-4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
-5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
-
-Training should take several hours. You will find checkpoints in `outputs/train/act_moss_test/checkpoints`.
-
-## Evaluate your policy
-
-You can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) but with a policy checkpoint as input. For instance, run this command to record 10 evaluation episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=moss \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/eval_act_moss_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=10 \
-  --control.push_to_hub=true \
-  --control.policy.path=outputs/train/act_moss_test/checkpoints/last/pretrained_model
-```
-
-As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
-1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_act_moss_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_moss_test`).
-2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_moss_test`).
-
-## More
-
-Follow this [previous tutorial](https://github.com/huggingface/lerobot/blob/main/examples/7_get_started_with_real_robot.md#4-train-a-policy-on-your-data) for a more in-depth tutorial on controlling real robots with LeRobot.
-
-If you have any question or need help, please reach out on Discord in the channel [`#moss-arm`](https://discord.com/channels/1216765309076115607/1275374638985252925).

examples/12_use_so101.md

@@ -1,709 +0,0 @@
-# Assemble and use SO-101
-
-In the steps below we explain how to assemble and use our flagship robot, the SO-101 with LeRobot 🤗.
-
-## Source the parts
-
-Follow this [README](https://github.com/TheRobotStudio/SO-ARM100). It contains the bill of materials, with a link to source the parts, as well as the instructions to 3D print the parts,
-and advice if it's your first time printing or if you don't own a 3D printer.
-
-Before assembling, you will first need to configure your motors. To this end, we provide a nice script, so let's first install LeRobot. After configuration, we will also guide you through assembly.
-
-## Install LeRobot
-
-> [!TIP]
-> We use the Command Prompt (cmd) quite a lot. If you are not comfortable using the cmd or want to brush up using the command line you can have a look here: [Command line crash course](https://developer.mozilla.org/en-US/docs/Learn_web_development/Getting_started/Environment_setup/Command_line)
-
-Download our source code:
-```bash
-git clone https://github.com/huggingface/lerobot.git
-cd lerobot
-```
-
-Create a virtual environment with Python 3.10 and activate it, e.g. with [`miniconda`](https://docs.anaconda.com/miniconda/install/#quick-command-line-install):
-```bash
-conda create -y -n lerobot python=3.10
-```
-Now restart the shell by running:
-
-##### Windows:
-```bash
-`source ~/.bashrc`
-```
-
-##### Mac:
-```bash
-`source ~/.bash_profile`
-```
-
-##### zshell:
-```bash
-`source ~/.zshrc`
-```
-
-Then activate your conda environment, you have to do this each time you open a shell to use lerobot:
-```bash
-conda activate lerobot
-```
-
-When using `miniconda`, install `ffmpeg` in your environment:
-```bash
-conda install ffmpeg -c conda-forge
-```
-
-> [!NOTE]
-> This usually installs `ffmpeg 7.X` for your platform compiled with the `libsvtav1` encoder. If `libsvtav1` is not supported (check supported encoders with `ffmpeg -encoders`), you can:
->  - _[On any platform]_ Explicitly install `ffmpeg 7.X` using:
->  ```bash
->  conda install ffmpeg=7.1.1 -c conda-forge
->  ```
->  - _[On Linux only]_ Install [ffmpeg build dependencies](https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#GettheDependencies) and [compile ffmpeg from source with libsvtav1](https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu#libsvtav1), and make sure you use the corresponding ffmpeg binary to your install with `which ffmpeg`.
-
-Install 🤗 LeRobot:
-```bash
-cd lerobot && pip install ".[feetech]"
-```
-
-> [!NOTE]
-> If you encounter build errors, you may need to install additional dependencies (`cmake`, `build-essential`, and `ffmpeg libs`). On Linux, run: `sudo apt-get install cmake build-essential python3-dev pkg-config libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libswscale-dev libswresample-dev libavfilter-dev pkg-config`. For other systems, see: [Compiling PyAV](https://pyav.org/docs/develop/overview/installation.html#bring-your-own-ffmpeg)
-
-
-## Configure motors
-
-To configure the motors designate one bus servo adapter and 6 motors for your leader arm, and similarly the other bus servo adapter and 6 motors for the follower arm. It's convenient to label them and write on each motor if it's for the follower `F` or for the leader `L` and it's ID from 1 to 6.
-
-You now should plug the 5V or 12V power supply to the motor bus. 5V for the STS3215 7.4V motors and 12V for the STS3215 12V motors. Note that the leader arm always uses the 7.4V motors, so watch out that you plug in the right power supply if you have 12V and 7.4V motors, otherwise you might burn your motors! Now, connect the motor bus to your computer via USB. Note that the USB doesn't provide any power, and both the power supply and USB have to be plugged in.
-
-### Find the USB ports associated to each arm
-
-To find the port for each bus servo adapter, run this script:
-```bash
-python lerobot/scripts/find_motors_bus_port.py
-```
-#### Example outputs of script
-
-##### Mac:
-Example output leader arm's port: `/dev/tty.usbmodem575E0031751`
-
-```bash
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this MotorsBus is /dev/tty.usbmodem575E0031751
-Reconnect the usb cable.
-```
-
-Example output follower arm port: `/dev/tty.usbmodem575E0032081`
-
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this MotorsBus is /dev/tty.usbmodem575E0032081
-Reconnect the usb cable.
-```
-
-##### Linux:
-On Linux, you might need to give access to the USB ports by running:
-```bash
-sudo chmod 666 /dev/ttyACM0
-sudo chmod 666 /dev/ttyACM1
-```
-
-Example output leader arm port: `/dev/ttyACM0`
-
-```bash
-Finding all available ports for the MotorBus.
-['/dev/ttyACM0', '/dev/ttyACM1']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this MotorsBus is /dev/ttyACM0
-Reconnect the usb cable.
-```
-
-Example output follower arm port: `/dev/ttyACM1`
-
-```
-Finding all available ports for the MotorBus.
-['/dev/ttyACM0', '/dev/ttyACM1']
-Remove the usb cable from your MotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this MotorsBus is /dev/ttyACM1
-Reconnect the usb cable.
-```
-
-#### Update config file
-
-Now that you have your ports, update the **port** default values of [`SO101RobotConfig`](https://github.com/huggingface/lerobot/blob/main/lerobot/common/robot_devices/robots/configs.py).
-You will find something a class called `so101` where you can update the `port` values with your actual motor ports:
-```python
-@RobotConfig.register_subclass("so101")
-@dataclass
-class So101RobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/so101"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",  <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-```
-
-Here is a video of the process:
-
-<video controls width="640" src="https://github.com/user-attachments/assets/fc45d756-31bb-4a61-b973-a87d633d08a7" type="video/mp4"></video>
-
-### Set motor IDs
-
-Now we need to set the motor ID for each motor. Plug your motor in only one of the two ports of the motor bus and run this script to set its ID to 1. Replace the text after --port to the corresponding control board port.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 1
-```
-
-Then unplug your motor and plug the second motor and set its ID to 2.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 2
-```
-
-Redo this process for all your motors until ID 6. Do the same for the 6 motors of the leader arm, but make sure to change the power supply if you use motors with different voltage.
-
-Here is a video of the process:
-
-<video controls width="640" src="https://github.com/user-attachments/assets/b31c115f-e706-4dcd-b7f1-4535da62416d" type="video/mp4"></video>
-
-## Step-by-Step Assembly Instructions
-
-The follower arm uses 6x STS3215 motors with 1/345 gearing. The leader however uses three differently geared motors to make sure it can both sustain its own weight and it can be moved without requiring much force. Which motor is needed for which joint is shown in table below.
-
-| Leader-Arm Axis | Motor | Gear Ratio |
-|-----------------|:-------:|:----------:|
-| Base / Shoulder Yaw | 1 | 1 / 191 |
-| Shoulder Pitch      | 2 | 1 / 345 |
-| Elbow               | 3 | 1 / 191 |
-| Wrist Roll          | 4 | 1 / 147 |
-| Wrist Pitch         | 5 | 1 / 147 |
-| Gripper             | 6 | 1 / 147 |
-
-
-### Clean Parts
-Remove all support material from the 3D-printed parts.
-
-### Joint 1
-
-- Place the first motor into the base.
-- Fasten the motor with 4 M2x6mm screws (smallest screws). Two from the top and two from bottom.
-- Slide over the first motor holder and fasten it using two M2x6mm screws (one on each side).
-- Install both motor horns, securing the top horn with a M3x6mm screw.
-- Attach the shoulder part.
-- Tighten the shoulder part with 4 M3x6mm screws on top and 4 M3x6mm screws on the bottom
-- Add the shoulder motor holder.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/b0ee9dee-a2d0-445b-8489-02ebecb3d639" type="video/mp4"></video>
-
-### Joint 2
-
-- Slide the second motor in from the top.
-- Fasten the second motor with 4 M2x6mm screws.
-- Attach both motor horns to motor 2, again use the M3x6mm horn screw.
-- Attach the upper arm with 4 M3x6mm screws on each side.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/32453dc2-5006-4140-9f56-f0d78eae5155" type="video/mp4"></video>
-
-### Joint 3
-
-- Insert motor 3 and fasten using 4 M2x6mm screws
-- Attach both motor horns to motor 3 and secure one again with a M3x6mm horn screw.
-- Connect the forearm to motor 3 using 4 M3x6mm screws on each side.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/7384b9a7-a946-440c-b292-91391bcc4d6b" type="video/mp4"></video>
-
-### Joint 4
-
-- Slide over motor holder 4.
-- Slide in motor 4.
-- Fasten motor 4 with 4 M2x6mm screws and attach its motor horns, use a M3x6mm horn screw.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/dca78ad0-7c36-4bdf-8162-c9ac42a1506f" type="video/mp4"></video>
-
-### Joint 5
-
-- Insert motor 5 into the wrist holder and secure it with 2 M2x6mm front screws.
-- Install only one motor horn on the wrist motor and secure it with a M3x6mm horn screw.
-- Secure the wrist to motor 4 using 4 M3x6mm screws on both sides.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/55f5d245-976d-49ff-8b4a-59843c441b12" type="video/mp4"></video>
-
-### Gripper / Handle
-
-#### Follower:
-
-- Attach the gripper to motor 5, attach it to the motor horn on the wrist using 4 M3x6mm screws.
-- Insert the gripper motor and secure it with 2 M2x6mm screws on each side.
-- Attach the motor horns and again use a M3x6mm horn screw.
-- Install the gripper claw and secure it with 4 M3x6mm screws on both sides.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/6f766aa9-cfae-4388-89e7-0247f198c086" type="video/mp4"></video>
-
-#### Leader:
-
-- Mount the leader holder onto the wrist and secure it with 4 M3x6mm screws.
-- Attach the handle to motor 5 using 1 M2x6mm screw.
-- Insert the gripper motor, secure it with 2 M2x6mm screws on each side, attach a motor horn using a M3x6mm horn screw.
-- Attach the follower trigger with 4 M3x6mm screws.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/1308c93d-2ef1-4560-8e93-a3812568a202" type="video/mp4"></video>
-
-##### Wiring
-
-- Attach the motor controller on the back.
-- Then insert all wires, use the wire guides everywhere to make sure the wires don't unplug themself and stay in place.
-
-<video controls width="640" src="https://github.com/user-attachments/assets/4c2cacfd-9276-4ee4-8bf2-ba2492667b78" type="video/mp4"></video>
-
-## Calibrate
-
-Next, you'll need to calibrate your SO-101 robot to ensure that the leader and follower arms have the same position values when they are in the same physical position.
-The calibration process is very important because it allows a neural network trained on one SO-101 robot to work on another.
-
-#### Manual calibration of follower arm
-
-You will need to move the follower arm to these positions sequentially, note that the rotated position is on the right side of the robot and you have to open the gripper fully.
-
-| 1. Middle position | 2. Zero position                                                                                                                                       | 3. Rotated position                                                                                                                                             | 4. Rest position                                                                                                                                       |
-| ------------ |------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="../media/so101/follower_middle.webp?raw=true" alt="SO-101 leader arm middle position" title="SO-101 leader arm middle position" style="width:100%;"> | <img src="../media/so101/follower_zero.webp?raw=true" alt="SO-101 leader arm zero position" title="SO-101 leader arm zero position" style="width:100%;"> | <img src="../media/so101/follower_rotated.webp?raw=true" alt="SO-101 leader arm rotated position" title="SO-101 leader arm rotated position" style="width:100%;"> | <img src="../media/so101/follower_rest.webp?raw=true" alt="SO-101 leader arm rest position" title="SO-101 leader arm rest position" style="width:100%;"> |
-
-Make sure both arms are connected and run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_follower"]'
-```
-
-#### Manual calibration of leader arm
-You will also need to move the leader arm to these positions sequentially:
-
-| 1. Middle position | 2. Zero position                                                                                                                                       | 3. Rotated position                                                                                                                                             | 4. Rest position                                                                                                                                       |
-| ------------ |------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| <img src="../media/so101/leader_middle.webp?raw=true" alt="SO-101 leader arm middle position" title="SO-101 leader arm middle position" style="width:100%;"> | <img src="../media/so101/leader_zero.webp?raw=true" alt="SO-101 leader arm zero position" title="SO-101 leader arm zero position" style="width:100%;"> | <img src="../media/so101/leader_rotated.webp?raw=true" alt="SO-101 leader arm rotated position" title="SO-101 leader arm rotated position" style="width:100%;"> | <img src="../media/so101/leader_rest.webp?raw=true" alt="SO-101 leader arm rest position" title="SO-101 leader arm rest position" style="width:100%;"> |
-
-Run this script to launch manual calibration:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --robot.cameras='{}' \
-  --control.type=calibrate \
-  --control.arms='["main_leader"]'
-```
-## Control your robot
-
-Congrats 🎉, your robot is all set to learn a task on its own. Next we will explain you how to train a neural network to autonomously control a real robot.
-
-**You'll learn to:**
-1. How to record and visualize your dataset.
-2. How to train a policy using your data and prepare it for evaluation.
-3. How to evaluate your policy and visualize the results.
-
-By following these steps, you'll be able to replicate tasks like picking up a Lego block and placing it in a bin with a high success rate, as demonstrated in [this video](https://x.com/RemiCadene/status/1814680760592572934).
-
-This tutorial is specifically made for the affordable [SO-101](https://github.com/TheRobotStudio/SO-ARM100) robot, but it contains additional information to be easily adapted to various types of robots like [Aloha bimanual robot](https://aloha-2.github.io) by changing some configurations. The SO-101 consists of a leader arm and a follower arm, each with 6 motors. It can work with one or several cameras to record the scene, which serve as visual sensors for the robot.
-
-During the data collection phase, you will control the follower arm by moving the leader arm. This process is known as "teleoperation." This technique is used to collect robot trajectories. Afterward, you'll train a neural network to imitate these trajectories and deploy the network to enable your robot to operate autonomously.
-
-If you encounter any issues at any step of the tutorial, feel free to seek help on [Discord](https://discord.com/invite/s3KuuzsPFb) or don't hesitate to iterate with us on the tutorial by creating issues or pull requests.
-
-## Teleoperate
-
-Run this simple script to teleoperate your robot (it won't connect and display the cameras):
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --robot.cameras='{}' \
-  --control.type=teleoperate
-```
-
-The teleoperate command will automatically:
-1. Identify any missing calibrations and initiate the calibration procedure.
-2. Connect the robot and start teleoperation.
-
-## Setup Cameras
-
-To connect a camera you have three options:
-1. OpenCVCamera which allows us to use any camera: usb, realsense, laptop webcam
-2. iPhone camera with MacOS
-3. Phone camera on Linux
-
-### Use OpenCVCamera
-
-The [`OpenCVCamera`](../lerobot/common/robot_devices/cameras/opencv.py) class allows you to efficiently record frames from most cameras using the [`opencv2`](https://docs.opencv.org) library.  For more details on compatibility, see [Video I/O with OpenCV Overview](https://docs.opencv.org/4.x/d0/da7/videoio_overview.html).
-
-To instantiate an [`OpenCVCamera`](../lerobot/common/robot_devices/cameras/opencv.py), you need a camera index (e.g. `OpenCVCamera(camera_index=0)`). When you only have one camera like a webcam of a laptop, the camera index is usually `0` but it might differ, and the camera index might change if you reboot your computer or re-plug your camera. This behavior depends on your operating system.
-
-To find the camera indices, run the following utility script, which will save a few frames from each detected camera:
-```bash
-python lerobot/common/robot_devices/cameras/opencv.py \
-    --images-dir outputs/images_from_opencv_cameras
-```
-
-The output will look something like this if you have two cameras connected:
-```
-Mac or Windows detected. Finding available camera indices through scanning all indices from 0 to 60
-[...]
-Camera found at index 0
-Camera found at index 1
-[...]
-Connecting cameras
-OpenCVCamera(0, fps=30.0, width=1920.0, height=1080.0, color_mode=rgb)
-OpenCVCamera(1, fps=24.0, width=1920.0, height=1080.0, color_mode=rgb)
-Saving images to outputs/images_from_opencv_cameras
-Frame: 0000 Latency (ms): 39.52
-[...]
-Frame: 0046 Latency (ms): 40.07
-Images have been saved to outputs/images_from_opencv_cameras
-```
-
-Check the saved images in `outputs/images_from_opencv_cameras` to identify which camera index corresponds to which physical camera (e.g. `0` for `camera_00` or `1` for `camera_01`):
-```
-camera_00_frame_000000.png
-[...]
-camera_00_frame_000047.png
-camera_01_frame_000000.png
-[...]
-camera_01_frame_000047.png
-```
-
-Note: Some cameras may take a few seconds to warm up, and the first frame might be black or green.
-
-Now that you have the camera indexes, you should change them in the config. You can also change the fps, width or height of the camera.
-
-The camera config is defined per robot, can be found here [`RobotConfig`](https://github.com/huggingface/lerobot/blob/main/lerobot/common/robot_devices/robots/configs.py) and looks like this:
-```python
-cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "wrist": OpenCVCameraConfig(
-                camera_index=0,  <-- UPDATE HERE
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "base": OpenCVCameraConfig(
-                camera_index=1,   <-- UPDATE HERE
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-```
-
-### Use your phone
-#### Mac:
-
-To use your iPhone as a camera on macOS, enable the Continuity Camera feature:
-- Ensure your Mac is running macOS 13 or later, and your iPhone is on iOS 16 or later.
-- Sign in both devices with the same Apple ID.
-- Connect your devices with a USB cable or turn on Wi-Fi and Bluetooth for a wireless connection.
-
-For more details, visit [Apple support](https://support.apple.com/en-gb/guide/mac-help/mchl77879b8a/mac).
-
-Your iPhone should be detected automatically when running the camera setup script in the next section.
-
-#### Linux:
-
-If you want to use your phone as a camera on Linux, follow these steps to set up a virtual camera
-
-1. *Install `v4l2loopback-dkms` and `v4l-utils`*. Those packages are required to create virtual camera devices (`v4l2loopback`) and verify their settings with the `v4l2-ctl` utility from `v4l-utils`. Install them using:
-```python
-sudo apt install v4l2loopback-dkms v4l-utils
-```
-2. *Install [DroidCam](https://droidcam.app) on your phone*. This app is available for both iOS and Android.
-3. *Install [OBS Studio](https://obsproject.com)*. This software will help you manage the camera feed. Install it using [Flatpak](https://flatpak.org):
-```python
-flatpak install flathub com.obsproject.Studio
-```
-4. *Install the DroidCam OBS plugin*. This plugin integrates DroidCam with OBS Studio. Install it with:
-```python
-flatpak install flathub com.obsproject.Studio.Plugin.DroidCam
-```
-5. *Start OBS Studio*. Launch with:
-```python
-flatpak run com.obsproject.Studio
-```
-6. *Add your phone as a source*. Follow the instructions [here](https://droidcam.app/obs/usage). Be sure to set the resolution to `640x480`.
-7. *Adjust resolution settings*. In OBS Studio, go to `File > Settings > Video`. Change the `Base(Canvas) Resolution` and the `Output(Scaled) Resolution` to `640x480` by manually typing it in.
-8. *Start virtual camera*. In OBS Studio, follow the instructions [here](https://obsproject.com/kb/virtual-camera-guide).
-9. *Verify the virtual camera setup*. Use `v4l2-ctl` to list the devices:
-```python
-v4l2-ctl --list-devices
-```
-You should see an entry like:
-```
-VirtualCam (platform:v4l2loopback-000):
-/dev/video1
-```
-10. *Check the camera resolution*. Use `v4l2-ctl` to ensure that the virtual camera output resolution is `640x480`. Change `/dev/video1` to the port of your virtual camera from the output of `v4l2-ctl --list-devices`.
-```python
-v4l2-ctl -d /dev/video1 --get-fmt-video
-```
-You should see an entry like:
-```
->>> Format Video Capture:
->>> Width/Height      : 640/480
->>> Pixel Format      : 'YUYV' (YUYV 4:2:2)
-```
-
-Troubleshooting: If the resolution is not correct you will have to delete the Virtual Camera port and try again as it cannot be changed.
-
-If everything is set up correctly, you can proceed with the rest of the tutorial.
-
-### Add wrist camera
-If you have an additional camera you can add a wrist camera to the SO101. There are already many premade wrist camera holders that you can find in the SO101 repo: [Wrist camera's](https://github.com/TheRobotStudio/SO-ARM100#wrist-cameras)
-
-## Teleoperate with cameras
-
-We can now teleoperate again while at the same time visualizing the camera's and joint positions with `rerun`.
-
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=teleoperate \
-  --control.display_data=true
-```
-
-## Record a dataset
-
-Once you're familiar with teleoperation, you can record your first dataset with SO-100.
-
-We use the Hugging Face hub features for uploading your dataset. If you haven't previously used the Hub, make sure you can login via the cli using a write-access token, this token can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens).
-
-Add your token to the cli by running this command:
-```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
-```
-
-Then store your Hugging Face repository name in a variable:
-```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
-echo $HF_USER
-```
-
-Now you can record a dataset, to record 2 episodes and upload your dataset to the hub execute this command:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/so101_test \
-  --control.tags='["so101","tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=2 \
-  --control.display_data=true \
-  --control.push_to_hub=true
-```
-
-You will see a lot of lines appearing like this one:
-```
-INFO 2024-08-10 15:02:58 ol_robot.py:219 dt:33.34 (30.0hz) dtRlead: 5.06 (197.5hz) dtWfoll: 0.25 (3963.7hz) dtRfoll: 6.22 (160.7hz) dtRlaptop: 32.57 (30.7hz) dtRphone: 33.84 (29.5hz)
-```
-It contains:
-- `2024-08-10 15:02:58` which is the date and time of the call to the print function,
-- `ol_robot.py:219` which is the end of the file name and the line number where the print function is called  (`lerobot/scripts/control_robot.py` line `219`).
-- `dt:33.34 (30.0hz)` which is the "delta time" or the number of milliseconds spent between the previous call to `robot.teleop_step(record_data=True)` and the current one, associated with the frequency (33.34 ms equals 30.0 Hz) ; note that we use `--fps 30` so we expect 30.0 Hz ; when a step takes more time, the line appears in yellow.
-- `dtRlead: 5.06 (197.5hz)` which is the delta time of reading the present position of the leader arm.
-- `dtWfoll: 0.25 (3963.7hz)` which is the delta time of writing the goal position on the follower arm ; writing is asynchronous so it takes less time than reading.
-- `dtRfoll: 6.22 (160.7hz)` which is the delta time of reading the present position on the follower arm.
-- `dtRlaptop:32.57 (30.7hz) ` which is the delta time of capturing an image from the laptop camera in the thread running asynchronously.
-- `dtRphone:33.84 (29.5hz)` which is the delta time of capturing an image from the phone camera in the thread running asynchronously.
-
-#### Dataset upload
-Locally your dataset is stored in this folder: `~/.cache/huggingface/lerobot/{repo-id}` (e.g. `data/cadene/so101_test`). At the end of data recording, your dataset will be uploaded on your Hugging Face page (e.g. https://huggingface.co/datasets/cadene/so101_test) that you can obtain by running:
-```bash
-echo https://huggingface.co/datasets/${HF_USER}/so101_test
-```
-Your dataset will be automatically tagged with `LeRobot` for the community to find it easily, and you can also add custom tags (in this case `tutorial` for example).
-
-You can look for other LeRobot datasets on the hub by searching for `LeRobot` [tags](https://huggingface.co/datasets?other=LeRobot).
-
-#### Record function
-
-The `record` function provides a suite of tools for capturing and managing data during robot operation:
-1. Set the flow of data recording using command line arguments:
-   - `--control.warmup_time_s=10` defines the number of seconds before starting data collection. It allows the robot devices to warmup and synchronize (10 seconds by default).
-   - `--control.episode_time_s=60` defines the number of seconds for data recording for each episode (60 seconds by default).
-   - `--control.reset_time_s=60` defines the number of seconds for resetting the environment after each episode (60 seconds by default).
-   - `--control.num_episodes=50` defines the number of episodes to record (50 by default).
-2. Control the flow during data recording using keyboard keys:
-   - Press right arrow `->` at any time during episode recording to early stop and go to resetting. Same during resetting, to early stop and to go to the next episode recording.
-   - Press left arrow `<-` at any time during episode recording or resetting to early stop, cancel the current episode, and re-record it.
-   - Press escape `ESC` at any time during episode recording to end the session early and go straight to video encoding and dataset uploading.
-3. Checkpoints are done set during recording, so if any issue occurs, you can resume recording by re-running the same command again with `--control.resume=true`. You will need to manually delete the dataset directory if you want to start recording from scratch.
-
-#### Tips for gathering data
-
-Once you're comfortable with data recording, you can create a larger dataset for training. A good starting task is grasping an object at different locations and placing it in a bin. We suggest recording at least 50 episodes, with 10 episodes per location. Keep the cameras fixed and maintain consistent grasping behavior throughout the recordings. Also make sure the object you are manipulating is visible on the camera's. A good rule of thumb is you should be able to do the task yourself by only looking at the camera images.
-
-In the following sections, you’ll train your neural network. After achieving reliable grasping performance, you can start introducing more variations during data collection, such as additional grasp locations, different grasping techniques, and altering camera positions.
-
-Avoid adding too much variation too quickly, as it may hinder your results.
-
-#### Troubleshooting:
-- On Linux, if the left and right arrow keys and escape key don't have any effect during data recording, make sure you've set the `$DISPLAY` environment variable. See [pynput limitations](https://pynput.readthedocs.io/en/latest/limitations.html#linux).
-
-## Visualize a dataset
-
-If you uploaded your dataset to the hub with `--control.push_to_hub=true`, you can [visualize your dataset online](https://huggingface.co/spaces/lerobot/visualize_dataset) by copy pasting your repo id given by:
-```bash
-echo ${HF_USER}/so101_test
-```
-
-If you didn't upload with `--control.push_to_hub=false`, you can visualize it locally with (via a window in the browser `http://127.0.0.1:9090` with the visualization tool):
-```bash
-python lerobot/scripts/visualize_dataset_html.py \
-  --repo-id ${HF_USER}/so101_test \
-  --local-files-only 1
-```
-
-This will launch a local web server that looks like this:
-
-<div style="text-align:center;">
-  <img src="../media/tutorial/visualize_dataset_html.webp?raw=true" alt="Koch v1.1 leader and follower arms" title="Koch v1.1 leader and follower arms" width="100%"></img>
-</div>
-
-## Replay an episode
-
-A useful feature is the `replay` function, which allows to replay on your robot any episode that you've recorded or episodes from any dataset out there. This function helps you test the repeatability of your robot's actions and assess transferability across robots of the same model.
-
-You can replay the first episode on your robot with:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=replay \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/so101_test \
-  --control.episode=0
-```
-
-Your robot should replicate movements similar to those you recorded. For example, check out [this video](https://x.com/RemiCadene/status/1793654950905680090) where we use `replay` on a Aloha robot from [Trossen Robotics](https://www.trossenrobotics.com).
-
-## Train a policy
-
-To train a policy to control your robot, use the [`python lerobot/scripts/train.py`](../lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/so101_test \
-  --policy.type=act \
-  --output_dir=outputs/train/act_so101_test \
-  --job_name=act_so101_test \
-  --policy.device=cuda \
-  --wandb.enable=true
-```
-
-Let's explain the command:
-1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/so101_test`.
-2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../lerobot/common/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor states, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
-4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
-5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
-
-Training should take several hours. You will find checkpoints in `outputs/train/act_so101_test/checkpoints`.
-
-To resume training from a checkpoint, below is an example command to resume from `last` checkpoint of the `act_so101_test` policy:
-```bash
-python lerobot/scripts/train.py \
-  --config_path=outputs/train/act_so101_test/checkpoints/last/pretrained_model/train_config.json \
-  --resume=true
-```
-
-#### Upload policy checkpoints
-
-Once training is done, upload the latest checkpoint with:
-```bash
-huggingface-cli upload ${HF_USER}/act_so101_test \
-  outputs/train/act_so101_test/checkpoints/last/pretrained_model
-```
-
-You can also upload intermediate checkpoints with:
-```bash
-CKPT=010000
-huggingface-cli upload ${HF_USER}/act_so101_test${CKPT} \
-  outputs/train/act_so101_test/checkpoints/${CKPT}/pretrained_model
-```
-
-## Evaluate your policy
-
-You can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) but with a policy checkpoint as input. For instance, run this command to record 10 evaluation episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=so101 \
-  --control.type=record \
-  --control.fps=30 \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.repo_id=${HF_USER}/eval_act_so101_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=10 \
-  --control.push_to_hub=true \
-  --control.policy.path=outputs/train/act_so101_test/checkpoints/last/pretrained_model
-```
-
-As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
-1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_act_so101_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_so101_test`).
-2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_so101_test`).

examples/1_load_lerobot_dataset.py

@@ -32,7 +32,7 @@ import torch
 from huggingface_hub import HfApi
 
 import lerobot
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
+from lerobot.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
 
 # We ported a number of existing datasets ourselves, use this to see the list:
 print("List of available datasets:")

examples/2_evaluate_pretrained_policy.py

@@ -30,7 +30,7 @@ import imageio
 import numpy
 import torch
 
-from lerobot.common.policies.diffusion.modeling_diffusion import DiffusionPolicy
+from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
 
 # Create a directory to store the video of the evaluation
 output_directory = Path("outputs/eval/example_pusht_diffusion")

examples/3_train_policy.py

@@ -22,11 +22,11 @@ from pathlib import Path
 
 import torch
 
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
-from lerobot.common.datasets.utils import dataset_to_policy_features
-from lerobot.common.policies.diffusion.configuration_diffusion import DiffusionConfig
-from lerobot.common.policies.diffusion.modeling_diffusion import DiffusionPolicy
 from lerobot.configs.types import FeatureType
+from lerobot.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
+from lerobot.datasets.utils import dataset_to_policy_features
+from lerobot.policies.diffusion.configuration_diffusion import DiffusionConfig
+from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
 
 
 def main():

examples/4_train_policy_with_script.md

@@ -4,7 +4,7 @@ This tutorial will explain the training script, how to use it, and particularly
 
 ## The training script
 
-LeRobot offers a training script at [`lerobot/scripts/train.py`](../lerobot/scripts/train.py). At a high level it does the following:
+LeRobot offers a training script at [`lerobot/scripts/train.py`](../src/lerobot/scripts/train.py). At a high level it does the following:
 
 - Initialize/load a configuration for the following steps using.
 - Instantiates a dataset.
@@ -21,7 +21,7 @@ In the training script, the main function `train` expects a `TrainPipelineConfig
 def train(cfg: TrainPipelineConfig):
 ```
 
-You can inspect the `TrainPipelineConfig` defined in [`lerobot/configs/train.py`](../lerobot/configs/train.py) (which is heavily commented and meant to be a reference to understand any option)
+You can inspect the `TrainPipelineConfig` defined in [`lerobot/configs/train.py`](../src/lerobot/configs/train.py) (which is heavily commented and meant to be a reference to understand any option)
 
 When running the script, inputs for the command line are parsed thanks to the `@parser.wrap()` decorator and an instance of this class is automatically generated. Under the hood, this is done with [Draccus](https://github.com/dlwh/draccus) which is a tool dedicated to this purpose. If you're familiar with Hydra, Draccus can similarly load configurations from config files (.json, .yaml) and also override their values through command line inputs. Unlike Hydra, these configurations are pre-defined in the code through dataclasses rather than being defined entirely in config files. This allows for more rigorous serialization/deserialization, typing, and to manipulate configuration as objects directly in the code and not as dictionaries or namespaces (which enables nice features in an IDE such as autocomplete, jump-to-def, etc.)
 
@@ -50,9 +50,9 @@ By default, every field takes its default value specified in the dataclass. If a
 
 ## Specifying values from the CLI
 
-Let's say that we want to train [Diffusion Policy](../lerobot/common/policies/diffusion) on the [pusht](https://huggingface.co/datasets/lerobot/pusht) dataset, using the [gym_pusht](https://github.com/huggingface/gym-pusht) environment for evaluation. The command to do so would look like this:
+Let's say that we want to train [Diffusion Policy](../src/lerobot/policies/diffusion) on the [pusht](https://huggingface.co/datasets/lerobot/pusht) dataset, using the [gym_pusht](https://github.com/huggingface/gym-pusht) environment for evaluation. The command to do so would look like this:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --dataset.repo_id=lerobot/pusht \
     --policy.type=diffusion \
     --env.type=pusht
@@ -60,12 +60,12 @@ python lerobot/scripts/train.py \
 
 Let's break this down:
 - To specify the dataset, we just need to specify its `repo_id` on the hub which is the only required argument in the `DatasetConfig`. The rest of the fields have default values and in this case we are fine with those so we can just add the option `--dataset.repo_id=lerobot/pusht`.
-- To specify the policy, we can just select diffusion policy using `--policy` appended with `.type`. Here, `.type` is a special argument which allows us to select config classes inheriting from `draccus.ChoiceRegistry` and that have been decorated with the `register_subclass()` method. To have a better explanation of this feature, have a look at this [Draccus demo](https://github.com/dlwh/draccus?tab=readme-ov-file#more-flexible-configuration-with-choice-types). In our code, we use this mechanism mainly to select policies, environments, robots, and some other components like optimizers. The policies available to select are located in [lerobot/common/policies](../lerobot/common/policies)
-- Similarly, we select the environment with `--env.type=pusht`. The different environment configs are available in [`lerobot/common/envs/configs.py`](../lerobot/common/envs/configs.py)
+- To specify the policy, we can just select diffusion policy using `--policy` appended with `.type`. Here, `.type` is a special argument which allows us to select config classes inheriting from `draccus.ChoiceRegistry` and that have been decorated with the `register_subclass()` method. To have a better explanation of this feature, have a look at this [Draccus demo](https://github.com/dlwh/draccus?tab=readme-ov-file#more-flexible-configuration-with-choice-types). In our code, we use this mechanism mainly to select policies, environments, robots, and some other components like optimizers. The policies available to select are located in [lerobot/policies](../src/lerobot/policies)
+- Similarly, we select the environment with `--env.type=pusht`. The different environment configs are available in [`lerobot/envs/configs.py`](../src/lerobot/envs/configs.py)
 
-Let's see another example. Let's say you've been training [ACT](../lerobot/common/policies/act) on [lerobot/aloha_sim_insertion_human](https://huggingface.co/datasets/lerobot/aloha_sim_insertion_human) using the [gym-aloha](https://github.com/huggingface/gym-aloha) environment for evaluation with:
+Let's see another example. Let's say you've been training [ACT](../src/lerobot/policies/act) on [lerobot/aloha_sim_insertion_human](https://huggingface.co/datasets/lerobot/aloha_sim_insertion_human) using the [gym-aloha](https://github.com/huggingface/gym-aloha) environment for evaluation with:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --policy.type=act \
     --dataset.repo_id=lerobot/aloha_sim_insertion_human \
     --env.type=aloha \
@@ -74,9 +74,9 @@ python lerobot/scripts/train.py \
 > Notice we added `--output_dir` to explicitly tell where to write outputs from this run (checkpoints, training state, configs etc.). This is not mandatory and if you don't specify it, a default directory will be created from the current date and time, env.type and policy.type. This will typically look like `outputs/train/2025-01-24/16-10-05_aloha_act`.
 
 We now want to train a different policy for aloha on another task. We'll change the dataset and use [lerobot/aloha_sim_transfer_cube_human](https://huggingface.co/datasets/lerobot/aloha_sim_transfer_cube_human) instead. Of course, we also need to change the task of the environment as well to match this other task.
-Looking at the [`AlohaEnv`](../lerobot/common/envs/configs.py) config, the task is `"AlohaInsertion-v0"` by default, which corresponds to the task we trained on in the command above. The [gym-aloha](https://github.com/huggingface/gym-aloha?tab=readme-ov-file#description) environment also has the `AlohaTransferCube-v0` task which corresponds to this other task we want to train on. Putting this together, we can train this new policy on this different task using:
+Looking at the [`AlohaEnv`](../src/lerobot/envs/configs.py) config, the task is `"AlohaInsertion-v0"` by default, which corresponds to the task we trained on in the command above. The [gym-aloha](https://github.com/huggingface/gym-aloha?tab=readme-ov-file#description) environment also has the `AlohaTransferCube-v0` task which corresponds to this other task we want to train on. Putting this together, we can train this new policy on this different task using:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --policy.type=act \
     --dataset.repo_id=lerobot/aloha_sim_transfer_cube_human \
     --env.type=aloha \
@@ -111,7 +111,7 @@ Now, let's assume that we want to reproduce the run just above. That run has pro
 
 We can then simply load the config values from this file using:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --config_path=outputs/train/act_aloha_transfer/checkpoints/last/pretrained_model/ \
     --output_dir=outputs/train/act_aloha_transfer_2
 ```
@@ -119,7 +119,7 @@ python lerobot/scripts/train.py \
 
 Similarly to Hydra, we can still override some parameters in the CLI if we want to, e.g.:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --config_path=outputs/train/act_aloha_transfer/checkpoints/last/pretrained_model/ \
     --output_dir=outputs/train/act_aloha_transfer_2
     --policy.n_action_steps=80
@@ -128,7 +128,7 @@ python lerobot/scripts/train.py \
 
 `--config_path` can also accept the repo_id of a repo on the hub that contains a `train_config.json` file, e.g. running:
 ```bash
-python lerobot/scripts/train.py --config_path=lerobot/diffusion_pusht
+python -m lerobot.scripts.train --config_path=lerobot/diffusion_pusht
 ```
 will start a training run with the same configuration used for training [lerobot/diffusion_pusht](https://huggingface.co/lerobot/diffusion_pusht)
 
@@ -139,7 +139,7 @@ Being able to resume a training run is important in case it crashed or aborted f
 
 Let's reuse the command from the previous run and add a few more options:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --policy.type=act \
     --dataset.repo_id=lerobot/aloha_sim_transfer_cube_human \
     --env.type=aloha \
@@ -155,7 +155,7 @@ INFO 2025-01-24 16:10:56 ts/train.py:263 Checkpoint policy after step 100
 ```
 Now let's simulate a crash by killing the process (hit `ctrl`+`c`). We can then simply resume this run from the last checkpoint available with:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --config_path=outputs/train/run_resumption/checkpoints/last/pretrained_model/ \
     --resume=true
 ```
@@ -164,7 +164,7 @@ You should see from the logging that your training picks up from where it left o
 Another reason for which you might want to resume a run is simply to extend training and add more training steps. The number of training steps is set by the option `--steps`, which is 100 000 by default.
 You could double the number of steps of the previous run with:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --config_path=outputs/train/run_resumption/checkpoints/last/pretrained_model/ \
     --resume=true \
     --steps=200000
@@ -195,7 +195,7 @@ In addition to the features currently in Draccus, we've added a special `.path`
 
 For example, we could fine-tune a [policy pre-trained on the aloha transfer task](https://huggingface.co/lerobot/act_aloha_sim_transfer_cube_human) on the aloha insertion task. We can achieve this with:
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --policy.path=lerobot/act_aloha_sim_transfer_cube_human \
     --dataset.repo_id=lerobot/aloha_sim_insertion_human \
     --env.type=aloha \
@@ -236,7 +236,7 @@ We'll summarize here the main use cases to remember from this tutorial.
 
 #### Train a policy from scratch – CLI
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --policy.type=act \  # <- select 'act' policy
     --env.type=pusht \  # <- select 'pusht' environment
     --dataset.repo_id=lerobot/pusht  # <- train on this dataset
@@ -244,14 +244,14 @@ python lerobot/scripts/train.py \
 
 #### Train a policy from scratch - config file + CLI
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --config_path=path/to/pretrained_model \  # <- can also be a repo_id
     --policy.n_action_steps=80  # <- you may still override values
 ```
 
 #### Resume/continue a training run
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --config_path=checkpoint/pretrained_model/ \
     --resume=true \
     --steps=200000  # <- you can change some training parameters
@@ -259,7 +259,7 @@ python lerobot/scripts/train.py \
 
 #### Fine-tuning
 ```bash
-python lerobot/scripts/train.py \
+python -m lerobot.scripts.train \
     --policy.path=lerobot/act_aloha_sim_transfer_cube_human \  # <- can also be a local path to a checkpoint
     --dataset.repo_id=lerobot/aloha_sim_insertion_human \
     --env.type=aloha \

examples/7_get_started_with_real_robot.md

@@ -1,998 +0,0 @@
-# Getting Started with Real-World Robots
-
-This tutorial will guide you through the process of setting up and training a neural network to autonomously control a real robot.
-
-**What You'll Learn:**
-1. How to order and assemble your robot.
-2. How to connect, configure, and calibrate your robot.
-3. How to record and visualize your dataset.
-4. How to train a policy using your data and prepare it for evaluation.
-5. How to evaluate your policy and visualize the results.
-
-By following these steps, you'll be able to replicate tasks like picking up a Lego block and placing it in a bin with a high success rate, as demonstrated in [this video](https://x.com/RemiCadene/status/1814680760592572934).
-
-This tutorial is specifically made for the affordable [Koch v1.1](https://github.com/jess-moss/koch-v1-1) robot, but it contains additional information to be easily adapted to various types of robots like [Aloha bimanual robot](https://aloha-2.github.io) by changing some configurations. The Koch v1.1 consists of a leader arm and a follower arm, each with 6 motors. It can work with one or several cameras to record the scene, which serve as visual sensors for the robot.
-
-During the data collection phase, you will control the follower arm by moving the leader arm. This process is known as "teleoperation." This technique is used to collect robot trajectories. Afterward, you'll train a neural network to imitate these trajectories and deploy the network to enable your robot to operate autonomously.
-
-If you encounter any issues at any step of the tutorial, feel free to seek help on [Discord](https://discord.com/invite/s3KuuzsPFb) or don't hesitate to iterate with us on the tutorial by creating issues or pull requests. Thanks!
-
-## 1. Order and Assemble your Koch v1.1
-
-Follow the sourcing and assembling instructions provided on the [Koch v1.1 Github page](https://github.com/jess-moss/koch-v1-1). This will guide you through setting up both the follower and leader arms, as shown in the image below.
-
-<div style="text-align:center;">
-  <img src="../media/tutorial/koch_v1_1_leader_follower.webp?raw=true" alt="Koch v1.1 leader and follower arms" title="Koch v1.1 leader and follower arms" width="50%">
-</div>
-
-For a visual walkthrough of the assembly process, you can refer to [this video tutorial](https://youtu.be/8nQIg9BwwTk).
-
-## 2. Configure motors, calibrate arms, teleoperate your Koch v1.1
-
-First, install the additional dependencies required for robots built with dynamixel motors like Koch v1.1 by running one of the following commands (make sure gcc is installed).
-
-Using `pip`:
-```bash
-pip install -e ".[dynamixel]"
-```
-
-Using `poetry`:
-```bash
-poetry sync --extras "dynamixel"
-```
-
-Using `uv`:
-```bash
-uv sync --extra "dynamixel"
-```
-
-You are now ready to plug the 5V power supply to the motor bus of the leader arm (the smaller one) since all its motors only require 5V.
-
-Then plug the 12V power supply to the motor bus of the follower arm. It has two motors that need 12V, and the rest will be powered with 5V through the voltage convertor.
-
-Finally, connect both arms to your computer via USB. Note that the USB doesn't provide any power, and both arms need to be plugged in with their associated power supply to be detected by your computer.
-
-Now you are ready to configure your motors for the first time, as detailed in the sections below. In the upcoming sections, you'll learn about our classes and functions by running some python code in an interactive session, or by copy-pasting it in a python file.
-
-If you have already configured your motors the first time, you can streamline the process by directly running the teleoperate script (which is detailed further in the tutorial):
-
-> **NOTE:** To visualize the data, enable `--control.display_data=true`. This streams the data using `rerun`.
-
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=koch \
-  --control.type=teleoperate
-```
-
-It will automatically:
-1. Identify any missing calibrations and initiate the calibration procedure.
-2. Connect the robot and start teleoperation.
-
-### a. Control your motors with DynamixelMotorsBus
-
-You can use the [`DynamixelMotorsBus`](../lerobot/common/robot_devices/motors/dynamixel.py) to communicate with the motors connected as a chain to the corresponding USB bus. This class leverages the Python [Dynamixel SDK](https://emanual.robotis.com/docs/en/software/dynamixel/dynamixel_sdk/sample_code/python_read_write_protocol_2_0/#python-read-write-protocol-20) to facilitate reading from and writing to the motors.
-
-**First Configuration of your motors**
-
-You will need to unplug each motor in turn and run a command the identify the motor. The motor will save its own identification, so you only need to do this once. Start by unplugging all of the motors.
-
-Do the Leader arm first, as all of its motors are of the same type. Plug in your first motor on your leader arm and run this script to set its ID to 1.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand dynamixel \
-  --model xl330-m288 \
-  --baudrate 1000000 \
-  --ID 1
-```
-
-Then unplug your first motor and plug the second motor and set its ID to 2.
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem58760432961 \
-  --brand dynamixel \
-  --model xl330-m288 \
-  --baudrate 1000000 \
-  --ID 2
-```
-
-Redo the process for all your motors until ID 6.
-
-The process for the follower arm is almost the same, but the follower arm has two types of motors. For the first two motors, make sure you set the model to `xl430-w250`. _Important: configuring follower motors requires plugging and unplugging power. Make sure you use the 5V power for the XL330s and the 12V power for the XL430s!_
-
-After all of your motors are configured properly, you're ready to plug them all together in a daisy-chain as shown in the original video.
-
-**Instantiate the DynamixelMotorsBus**
-
-To begin, create two instances of the  [`DynamixelMotorsBus`](../lerobot/common/robot_devices/motors/dynamixel.py), one for each arm, using their corresponding USB ports (e.g. `DynamixelMotorsBus(port="/dev/tty.usbmodem575E0031751"`).
-
-To find the correct ports for each arm, run the utility script twice:
-```bash
-python lerobot/scripts/find_motors_bus_port.py
-```
-
-Example output when identifying the leader arm's port (e.g., `/dev/tty.usbmodem575E0031751` on Mac, or possibly `/dev/ttyACM0` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-
-[...Disconnect leader arm and press Enter...]
-
-The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0031751
-Reconnect the usb cable.
-```
-
-Example output when identifying the follower arm's port (e.g., `/dev/tty.usbmodem575E0032081`, or possibly `/dev/ttyACM1` on Linux):
-```
-Finding all available ports for the MotorBus.
-['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-
-[...Disconnect follower arm and press Enter...]
-
-The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0032081
-Reconnect the usb cable.
-```
-
-Troubleshooting: On Linux, you might need to give access to the USB ports by running this command with your ports:
-```bash
-sudo chmod 666 /dev/tty.usbmodem575E0032081
-sudo chmod 666 /dev/tty.usbmodem575E0031751
-```
-
-*Listing and Configuring Motors*
-
-Next, you'll need to list the motors for each arm, including their name, index, and model. Initially, each motor is assigned the factory default index `1`. Since each motor requires a unique index to function correctly when connected in a chain on a common bus, you'll need to assign different indices. It's recommended to use an ascending index order, starting from `1` (e.g., `1, 2, 3, 4, 5, 6`). These indices will be saved in the persistent memory of each motor during the first connection.
-
-To assign indices to the motors, run this code in an interactive Python session. Replace the `port` values with the ones you identified earlier:
-```python
-from lerobot.common.robot_devices.motors.configs import DynamixelMotorsBusConfig
-from lerobot.common.robot_devices.motors.dynamixel import DynamixelMotorsBus
-
-leader_config = DynamixelMotorsBusConfig(
-    port="/dev/tty.usbmodem575E0031751",
-    motors={
-        # name: (index, model)
-        "shoulder_pan": (1, "xl330-m077"),
-        "shoulder_lift": (2, "xl330-m077"),
-        "elbow_flex": (3, "xl330-m077"),
-        "wrist_flex": (4, "xl330-m077"),
-        "wrist_roll": (5, "xl330-m077"),
-        "gripper": (6, "xl330-m077"),
-    },
-)
-
-follower_config = DynamixelMotorsBusConfig(
-    port="/dev/tty.usbmodem575E0032081",
-    motors={
-        # name: (index, model)
-        "shoulder_pan": (1, "xl430-w250"),
-        "shoulder_lift": (2, "xl430-w250"),
-        "elbow_flex": (3, "xl330-m288"),
-        "wrist_flex": (4, "xl330-m288"),
-        "wrist_roll": (5, "xl330-m288"),
-        "gripper": (6, "xl330-m288"),
-    },
-)
-
-leader_arm = DynamixelMotorsBus(leader_config)
-follower_arm = DynamixelMotorsBus(follower_config)
-```
-
-IMPORTANTLY: Now that you have your ports, update [`KochRobotConfig`](../lerobot/common/robot_devices/robots/configs.py). You will find something like:
-```python
-@RobotConfig.register_subclass("koch")
-@dataclass
-class KochRobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/koch"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0085511", <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl330-m077"],
-                    "shoulder_lift": [2, "xl330-m077"],
-                    "elbow_flex": [3, "xl330-m077"],
-                    "wrist_flex": [4, "xl330-m077"],
-                    "wrist_roll": [5, "xl330-m077"],
-                    "gripper": [6, "xl330-m077"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891", <-- UPDATE HERE
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl430-w250"],
-                    "shoulder_lift": [2, "xl430-w250"],
-                    "elbow_flex": [3, "xl330-m288"],
-                    "wrist_flex": [4, "xl330-m288"],
-                    "wrist_roll": [5, "xl330-m288"],
-                    "gripper": [6, "xl330-m288"],
-                },
-            ),
-        }
-    )
-```
-
-**Connect and Configure your Motors**
-
-Before you can start using your motors, you'll need to configure them to ensure proper communication. When you first connect the motors, the [`DynamixelMotorsBus`](../lerobot/common/robot_devices/motors/dynamixel.py) automatically detects any mismatch between the current motor indices (factory set to `1`) and the specified indices (e.g., `1, 2, 3, 4, 5, 6`). This triggers a configuration procedure that requires you to unplug the power cord and motors, then reconnect each motor sequentially, starting from the one closest to the bus.
-
-For a visual guide, refer to the [video tutorial of the configuration procedure](https://youtu.be/U78QQ9wCdpY).
-
-To connect and configure the leader arm, run the following code in the same Python interactive session as earlier in the tutorial:
-```python
-leader_arm.connect()
-```
-
-When you connect the leader arm for the first time, you might see an output similar to this:
-```
-Read failed due to communication error on port /dev/tty.usbmodem575E0032081 for group_key ID_shoulder_pan_shoulder_lift_elbow_flex_wrist_flex_wrist_roll_gripper: [TxRxResult] There is no status packet!
-
-/!\ A configuration issue has been detected with your motors:
-If this is the first time you are using these motors, press enter to configure your motors... but before verify that all the cables are connected the proper way. If you find an issue, before making a modification, kill the python process, unplug the power cord to not damage the motors, rewire correctly, then plug the power again and relaunch the script.
-
-Motor indices detected: {9600: [1]}
-
-1. Unplug the power cord
-2. Plug/unplug minimal number of cables to only have the first 1 motor(s) (['shoulder_pan']) connected.
-3. Re-plug the power cord
-Press Enter to continue...
-
-*Follow the procedure*
-
-Setting expected motor indices: [1, 2, 3, 4, 5, 6]
-```
-
-Once the leader arm is configured, repeat the process for the follower arm by running:
-```python
-follower_arm.connect()
-```
-
-Congratulations! Both arms are now properly configured and connected. You won't need to go through the configuration procedure again in the future.
-
-**Troubleshooting**:
-
-If the configuration process fails, you may need to do the configuration process via the Dynamixel Wizard.
-
-Known failure modes:
-- Calling `arm.connect()` raises `OSError: No motor found, but one new motor expected. Verify power cord is plugged in and retry` on Ubuntu 22.
-
-Steps:
-1. Visit https://emanual.robotis.com/docs/en/software/dynamixel/dynamixel_wizard2/#connect-dynamixel.
-2. Follow the software installation instructions in section 3 of the web page.
-3. Launch the software.
-4. Configure the device scanning options in the menu under `Tools` > `Options` > `Scan`. Check only Protocol 2.0, select only the USB port identifier of interest, select all baudrates, set the ID range to `[0, 10]`. _While this step was not strictly necessary, it greatly speeds up scanning_.
-5. For each motor in turn:
-    - Disconnect the power to the driver board.
-    - Connect **only** the motor of interest to the driver board, making sure to disconnect it from any other motors.
-    - Reconnect the power to the driver board.
-    - From the software menu select `Device` > `Scan` and let the scan run. A device should appear.
-    - If the device has an asterisk (*) near it, it means the firmware is indeed outdated. From the software menu, select `Tools` > `Firmware Update`. Follow the prompts.
-    - The main panel should have table with various parameters of the device (refer to the web page, section 5). Select the row with `ID`, and then set the desired ID on the bottom right panel by selecting and clicking `Save`.
-    - Just like you did with the ID, also set the `Baud Rate` to 1 Mbps.
-6. Check everything has been done right:
-   - Rewire the arms in their final configuration and power both of them.
-   - Scan for devices. All 12 motors should appear.
-   - Select the motors one by one and move the arm. Check that the graphical indicator near the top right shows the movement.
-
-** There is a common issue with the Dynamixel XL430-W250 motors where the motors become undiscoverable after upgrading their firmware from Mac and Windows Dynamixel Wizard2 applications.  When this occurs, it is required to do a firmware recovery (Select `DYNAMIXEL Firmware Recovery` and follow the prompts).   There are two known workarounds to conduct this firmware reset:
-  1) Install the Dynamixel Wizard on a linux machine and complete the firmware recovery
-  2) Use the Dynamixel U2D2 in order to perform the reset with Windows or Mac.  This U2D2 can be purchased [here](https://www.robotis.us/u2d2/).
-  For either solution, open DYNAMIXEL Wizard 2.0 and select the appropriate port. You will likely be unable to see the motor in the GUI at this time. Select `Firmware Recovery`, carefully choose the correct model, and wait for the process to complete. Finally, re-scan to confirm the firmware recovery was successful.
-
-**Read and Write with DynamixelMotorsBus**
-
-To get familiar with how `DynamixelMotorsBus` communicates with the motors, you can start by reading data from them. Copy past this code in the same interactive python session:
-```python
-leader_pos = leader_arm.read("Present_Position")
-follower_pos = follower_arm.read("Present_Position")
-print(leader_pos)
-print(follower_pos)
-```
-
-Expected output might look like:
-```
-array([2054,  523, 3071, 1831, 3049, 2441], dtype=int32)
-array([2003, 1601,   56, 2152, 3101, 2283], dtype=int32)
-```
-
-Try moving the arms to various positions and observe how the values change.
-
-Now let's try to enable torque in the follower arm by copy pasting this code:
-```python
-from lerobot.common.robot_devices.motors.dynamixel import TorqueMode
-
-follower_arm.write("Torque_Enable", TorqueMode.ENABLED.value)
-```
-
-With torque enabled, the follower arm will be locked in its current position. Do not attempt to manually move the arm while torque is enabled, as this could damage the motors.
-
-Now, to get more familiar with reading and writing, let's move the arm programmatically copy pasting the following example code:
-```python
-# Get the current position
-position = follower_arm.read("Present_Position")
-
-# Update first motor (shoulder_pan) position by +10 steps
-position[0] += 10
-follower_arm.write("Goal_Position", position)
-
-# Update all motors position by -30 steps
-position -= 30
-follower_arm.write("Goal_Position", position)
-
-# Update gripper by +30 steps
-position[-1] += 30
-follower_arm.write("Goal_Position", position[-1], "gripper")
-```
-
-When you're done playing, you can try to disable the torque, but make sure you hold your robot so that it doesn't fall:
-```python
-follower_arm.write("Torque_Enable", TorqueMode.DISABLED.value)
-```
-
-Finally, disconnect the arms:
-```python
-leader_arm.disconnect()
-follower_arm.disconnect()
-```
-
-Alternatively, you can unplug the power cord, which will automatically disable torque and disconnect the motors.
-
-*/!\ Warning*: These motors tend to overheat, especially under torque or if left plugged in for too long. Unplug after use.
-
-### b. Teleoperate your Koch v1.1 with ManipulatorRobot
-
-**Instantiate the ManipulatorRobot**
-
-Before you can teleoperate your robot, you need to instantiate the  [`ManipulatorRobot`](../lerobot/common/robot_devices/robots/manipulator.py) using the previously defined `leader_config` and `follower_config`.
-
-For the Koch v1.1 robot, we only have one leader, so we refer to it as `"main"` and define it as `leader_arms={"main": leader_config}`. We do the same for the follower arm. For other robots (like the Aloha), which may have two pairs of leader and follower arms, you would define them like this: `leader_arms={"left": left_leader_config, "right": right_leader_config},`. Same thing for the follower arms.
-
-
-Run the following code to instantiate your manipulator robot:
-```python
-from lerobot.common.robot_devices.robots.configs import KochRobotConfig
-from lerobot.common.robot_devices.robots.manipulator import ManipulatorRobot
-
-robot_config = KochRobotConfig(
-    leader_arms={"main": leader_config},
-    follower_arms={"main": follower_config},
-    cameras={},  # We don't use any camera for now
-)
-robot = ManipulatorRobot(robot_config)
-```
-
-The `KochRobotConfig` is used to set the associated settings and calibration process. For instance, we activate the torque of the gripper of the leader Koch v1.1 arm and position it at a 40 degree angle to use it as a trigger.
-
-For the [Aloha bimanual robot](https://aloha-2.github.io), we would use `AlohaRobotConfig` to set different settings such as a secondary ID for shadow joints (shoulder, elbow). Specific to Aloha, LeRobot comes with default calibration files stored in `.cache/calibration/aloha_default`. Assuming the motors have been properly assembled, no manual calibration step is expected for Aloha.
-
-**Calibrate and Connect the ManipulatorRobot**
-
-Next, you'll need to calibrate your Koch robot to ensure that the leader and follower arms have the same position values when they are in the same physical position. This calibration is essential because it allows a neural network trained on one Koch robot to work on another.
-
-When you connect your robot for the first time, the [`ManipulatorRobot`](../lerobot/common/robot_devices/robots/manipulator.py) will detect if the calibration file is missing and trigger the calibration procedure. During this process, you will be guided to move each arm to three different positions.
-
-Here are the positions you'll move the follower arm to:
-
-| 1. Zero position                                                                                                                                                  | 2. Rotated position                                                                                                                                                        | 3. Rest position                                                                                                                                                  |
-| ----------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| <img src="../media/koch/follower_zero.webp?raw=true" alt="Koch v1.1 follower arm zero position" title="Koch v1.1 follower arm zero position" style="width:100%;"> | <img src="../media/koch/follower_rotated.webp?raw=true" alt="Koch v1.1 follower arm rotated position" title="Koch v1.1 follower arm rotated position" style="width:100%;"> | <img src="../media/koch/follower_rest.webp?raw=true" alt="Koch v1.1 follower arm rest position" title="Koch v1.1 follower arm rest position" style="width:100%;"> |
-
-And here are the corresponding positions for the leader arm:
-
-| 1. Zero position                                                                                                                                            | 2. Rotated position                                                                                                                                                  | 3. Rest position                                                                                                                                            |
-| ----------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| <img src="../media/koch/leader_zero.webp?raw=true" alt="Koch v1.1 leader arm zero position" title="Koch v1.1 leader arm zero position" style="width:100%;"> | <img src="../media/koch/leader_rotated.webp?raw=true" alt="Koch v1.1 leader arm rotated position" title="Koch v1.1 leader arm rotated position" style="width:100%;"> | <img src="../media/koch/leader_rest.webp?raw=true" alt="Koch v1.1 leader arm rest position" title="Koch v1.1 leader arm rest position" style="width:100%;"> |
-
-You can watch a [video tutorial of the calibration procedure](https://youtu.be/8drnU9uRY24) for more details.
-
-During calibration, we count the number of full 360-degree rotations your motors have made since they were first used. That's why we ask you to move to this arbitrary "zero" position. We don't actually "set" the zero position, so you don't need to be accurate. After calculating these "offsets" to shift the motor values around 0, we need to assess the rotation direction of each motor, which might differ. That's why we ask you to rotate all motors to roughly 90 degrees, to measure if the values changed negatively or positively.
-
-Finally, the rest position ensures that the follower and leader arms are roughly aligned after calibration, preventing sudden movements that could damage the motors when starting teleoperation.
-
-Importantly, once calibrated, all Koch robots will move to the same positions (e.g. zero and rotated position) when commanded.
-
-Run the following code to calibrate and connect your robot:
-```python
-robot.connect()
-```
-
-The output will look like this:
-```
-Connecting main follower arm
-Connecting main leader arm
-
-Missing calibration file '.cache/calibration/koch/main_follower.json'
-Running calibration of koch main follower...
-Move arm to zero position
-[...]
-Move arm to rotated position
-[...]
-Move arm to rest position
-[...]
-Calibration is done! Saving calibration file '.cache/calibration/koch/main_follower.json'
-
-Missing calibration file '.cache/calibration/koch/main_leader.json'
-Running calibration of koch main leader...
-Move arm to zero position
-[...]
-Move arm to rotated position
-[...]
-Move arm to rest position
-[...]
-Calibration is done! Saving calibration file '.cache/calibration/koch/main_leader.json'
-```
-
-*Verifying Calibration*
-
-Once calibration is complete, you can check the positions of the leader and follower arms to ensure they match. If the calibration was successful, the positions should be very similar.
-
-Run this code to get the positions in degrees:
-```python
-leader_pos = robot.leader_arms["main"].read("Present_Position")
-follower_pos = robot.follower_arms["main"].read("Present_Position")
-
-print(leader_pos)
-print(follower_pos)
-```
-
-Example output:
-```
-array([-0.43945312, 133.94531, 179.82422, -18.984375, -1.9335938, 34.541016], dtype=float32)
-array([-0.58723712, 131.72314, 174.98743, -16.872612, 0.786213, 35.271973], dtype=float32)
-```
-
-These values are in degrees, which makes them easier to interpret and debug. The zero position used during calibration should roughly correspond to 0 degrees for each motor, and the rotated position should roughly correspond to 90 degrees for each motor.
-
-**Teleoperate your Koch v1.1**
-
-You can easily teleoperate your robot by reading the positions from the leader arm and sending them as goal positions to the follower arm.
-
-To teleoperate your robot for 30 seconds at a frequency of approximately 200Hz, run the following code:
-```python
-import tqdm
-seconds = 30
-frequency = 200
-for _ in tqdm.tqdm(range(seconds*frequency)):
-    leader_pos = robot.leader_arms["main"].read("Present_Position")
-    robot.follower_arms["main"].write("Goal_Position", leader_pos)
-```
-
-*Using `teleop_step` for Teleoperation*
-
-Alternatively, you can teleoperate the robot using the `teleop_step` method from [`ManipulatorRobot`](../lerobot/common/robot_devices/robots/manipulator.py).
-
-Run this code to teleoperate:
-```python
-for _ in tqdm.tqdm(range(seconds*frequency)):
-    robot.teleop_step()
-```
-
-*Recording data during Teleoperation*
-
-Teleoperation is particularly useful for recording data. You can use the `teleop_step(record_data=True)` to returns both the follower arm's position as `"observation.state"` and the leader arm's position as `"action"`. This function also converts the numpy arrays into PyTorch tensors. If you're working with a robot that has two leader and two follower arms (like the Aloha), the positions are concatenated.
-
-Run the following code to see how slowly moving the leader arm affects the observation and action:
-```python
-leader_pos = robot.leader_arms["main"].read("Present_Position")
-follower_pos = robot.follower_arms["main"].read("Present_Position")
-observation, action = robot.teleop_step(record_data=True)
-
-print(follower_pos)
-print(observation)
-print(leader_pos)
-print(action)
-```
-
-Expected output:
-```
-array([7.8223, 131.1328, 165.5859, -23.4668, -0.9668, 32.4316], dtype=float32)
-{'observation.state': tensor([7.8223, 131.1328, 165.5859, -23.4668, -0.9668, 32.4316])}
-array([3.4277, 134.1211, 179.8242, -18.5449, -1.5820, 34.7168], dtype=float32)
-{'action': tensor([3.4277, 134.1211, 179.8242, -18.5449, -1.5820, 34.7168])}
-```
-
-*Asynchronous Frame Recording*
-
-Additionally, `teleop_step` can asynchronously record frames from multiple cameras and include them in the observation dictionary as `"observation.images.CAMERA_NAME"`. This feature will be covered in more detail in the next section.
-
-*Disconnecting the Robot*
-
-When you're finished, make sure to disconnect your robot by running:
-```python
-robot.disconnect()
-```
-
-Alternatively, you can unplug the power cord, which will also disable torque.
-
-*/!\ Warning*: These motors tend to overheat, especially under torque or if left plugged in for too long. Unplug after use.
-
-### c. Add your cameras with OpenCVCamera
-
-**(Optional) Use your phone as camera on Linux**
-
-If you want to use your phone as a camera on Linux, follow these steps to set up a virtual camera
-
-1. *Install `v4l2loopback-dkms` and `v4l-utils`*. Those packages are required to create virtual camera devices (`v4l2loopback`) and verify their settings with the `v4l2-ctl` utility from `v4l-utils`. Install them using:
-```python
-sudo apt install v4l2loopback-dkms v4l-utils
-```
-2. *Install [DroidCam](https://droidcam.app) on your phone*. This app is available for both iOS and Android.
-3. *Install [OBS Studio](https://obsproject.com)*. This software will help you manage the camera feed. Install it using [Flatpak](https://flatpak.org):
-```python
-flatpak install flathub com.obsproject.Studio
-```
-4. *Install the DroidCam OBS plugin*. This plugin integrates DroidCam with OBS Studio. Install it with:
-```python
-flatpak install flathub com.obsproject.Studio.Plugin.DroidCam
-```
-5. *Start OBS Studio*. Launch with:
-```python
-flatpak run com.obsproject.Studio
-```
-6. *Add your phone as a source*. Follow the instructions [here](https://droidcam.app/obs/usage). Be sure to set the resolution to `640x480`.
-7. *Adjust resolution settings*. In OBS Studio, go to `File > Settings > Video`. Change the `Base(Canvas) Resolution` and the `Output(Scaled) Resolution` to `640x480` by manually typing it in.
-8. *Start virtual camera*. In OBS Studio, follow the instructions [here](https://obsproject.com/kb/virtual-camera-guide).
-9. *Verify the virtual camera setup*. Use `v4l2-ctl` to list the devices:
-```python
-v4l2-ctl --list-devices
-```
-You should see an entry like:
-```
-VirtualCam (platform:v4l2loopback-000):
-/dev/video1
-```
-10. *Check the camera resolution*. Use `v4l2-ctl` to ensure that the virtual camera output resolution is `640x480`. Change `/dev/video1` to the port of your virtual camera from the output of `v4l2-ctl --list-devices`.
-```python
-v4l2-ctl -d /dev/video1 --get-fmt-video
-```
-You should see an entry like:
-```
->>> Format Video Capture:
->>>	Width/Height      : 640/480
->>>	Pixel Format      : 'YUYV' (YUYV 4:2:2)
-```
-
-Troubleshooting: If the resolution is not correct you will have to delete the Virtual Camera port and try again as it cannot be changed.
-
-If everything is set up correctly, you can proceed with the rest of the tutorial.
-
-**(Optional) Use your iPhone as a camera on MacOS**
-
-To use your iPhone as a camera on macOS, enable the Continuity Camera feature:
-- Ensure your Mac is running macOS 13 or later, and your iPhone is on iOS 16 or later.
-- Sign in both devices with the same Apple ID.
-- Connect your devices with a USB cable or turn on Wi-Fi and Bluetooth for a wireless connection.
-
-For more details, visit [Apple support](https://support.apple.com/en-gb/guide/mac-help/mchl77879b8a/mac).
-
-Your iPhone should be detected automatically when running the camera setup script in the next section.
-
-**Instantiate an OpenCVCamera**
-
-The [`OpenCVCamera`](../lerobot/common/robot_devices/cameras/opencv.py) class allows you to efficiently record frames from most cameras using the [`opencv2`](https://docs.opencv.org) library.  For more details on compatibility, see [Video I/O with OpenCV Overview](https://docs.opencv.org/4.x/d0/da7/videoio_overview.html).
-
-To instantiate an [`OpenCVCamera`](../lerobot/common/robot_devices/cameras/opencv.py), you need a camera index (e.g. `OpenCVCamera(camera_index=0)`). When you only have one camera like a webcam of a laptop, the camera index is usually `0` but it might differ, and the camera index might change if you reboot your computer or re-plug your camera. This behavior depends on your operating system.
-
-To find the camera indices, run the following utility script, which will save a few frames from each detected camera:
-```bash
-python lerobot/common/robot_devices/cameras/opencv.py \
-    --images-dir outputs/images_from_opencv_cameras
-```
-
-The output will look something like this if you have two cameras connected:
-```
-Mac or Windows detected. Finding available camera indices through scanning all indices from 0 to 60
-[...]
-Camera found at index 0
-Camera found at index 1
-[...]
-Connecting cameras
-OpenCVCamera(0, fps=30.0, width=1920.0, height=1080.0, color_mode=rgb)
-OpenCVCamera(1, fps=24.0, width=1920.0, height=1080.0, color_mode=rgb)
-Saving images to outputs/images_from_opencv_cameras
-Frame: 0000	Latency (ms): 39.52
-[...]
-Frame: 0046	Latency (ms): 40.07
-Images have been saved to outputs/images_from_opencv_cameras
-```
-
-Check the saved images in `outputs/images_from_opencv_cameras` to identify which camera index corresponds to which physical camera (e.g. `0` for `camera_00` or `1` for `camera_01`):
-```
-camera_00_frame_000000.png
-[...]
-camera_00_frame_000047.png
-camera_01_frame_000000.png
-[...]
-camera_01_frame_000047.png
-```
-
-Note: Some cameras may take a few seconds to warm up, and the first frame might be black or green.
-
-Finally, run this code to instantiate and connect your camera:
-```python
-from lerobot.common.robot_devices.cameras.configs import OpenCVCameraConfig
-from lerobot.common.robot_devices.cameras.opencv import OpenCVCamera
-
-config = OpenCVCameraConfig(camera_index=0)
-camera = OpenCVCamera(config)
-camera.connect()
-color_image = camera.read()
-
-print(color_image.shape)
-print(color_image.dtype)
-```
-
-Expected output for a laptop camera on MacBookPro:
-```
-(1080, 1920, 3)
-uint8
-```
-
-Or like this if you followed our tutorial to set a virtual camera:
-```
-(480, 640, 3)
-uint8
-```
-
-With certain camera, you can also specify additional parameters like frame rate, resolution, and color mode during instantiation. For instance:
-```python
-config = OpenCVCameraConfig(camera_index=0, fps=30, width=640, height=480)
-```
-
-If the provided arguments are not compatible with the camera, an exception will be raised.
-
-*Disconnecting the camera*
-
-When you're done using the camera, disconnect it by running:
-```python
-camera.disconnect()
-```
-
-**Instantiate your robot with cameras**
-
-Additionally, you can set up your robot to work with your cameras.
-
-Modify the following Python code with the appropriate camera names and configurations:
-```python
-robot = ManipulatorRobot(
-    KochRobotConfig(
-        leader_arms={"main": leader_arm},
-        follower_arms={"main": follower_arm},
-        calibration_dir=".cache/calibration/koch",
-        cameras={
-            "laptop": OpenCVCameraConfig(0, fps=30, width=640, height=480),
-            "phone": OpenCVCameraConfig(1, fps=30, width=640, height=480),
-        },
-    )
-)
-robot.connect()
-```
-
-As a result, `teleop_step(record_data=True` will return a frame for each camera following the pytorch "channel first" convention but we keep images in `uint8` with pixels in range [0,255] to easily save them.
-
-Modify this code with the names of your cameras and run it:
-```python
-observation, action = robot.teleop_step(record_data=True)
-print(observation["observation.images.laptop"].shape)
-print(observation["observation.images.phone"].shape)
-print(observation["observation.images.laptop"].min().item())
-print(observation["observation.images.laptop"].max().item())
-```
-
-The output should look like this:
-```
-torch.Size([3, 480, 640])
-torch.Size([3, 480, 640])
-0
-255
-```
-
-### d. Use `control_robot.py` and our `teleoperate` function
-
-Instead of manually running the python code in a terminal window, you can use [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) to instantiate your robot by providing the robot configurations via command line and control your robot with various modes as explained next.
-
-Try running this code to teleoperate your robot (if you dont have a camera, keep reading):
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=koch \
-  --control.type=teleoperate
-```
-
-You will see a lot of lines appearing like this one:
-```
-INFO 2024-08-10 11:15:03 ol_robot.py:209 dt: 5.12 (195.1hz) dtRlead: 4.93 (203.0hz) dtWfoll: 0.19 (5239.0hz)
-```
-
-It contains
-- `2024-08-10 11:15:03` which is the date and time of the call to the print function.
-- `ol_robot.py:209` which is the end of the file name and the line number where the print function is called  (`lerobot/scripts/control_robot.py` line `209`).
-- `dt: 5.12 (195.1hz)` which is the "delta time" or the number of milliseconds spent between the previous call to `robot.teleop_step()` and the current one, associated with the frequency (5.12 ms equals 195.1 Hz) ; note that you can control the maximum frequency by adding fps as argument such as `--fps 30`.
-- `dtRlead: 4.93 (203.0hz)` which is the number of milliseconds it took to read the position of the leader arm using `leader_arm.read("Present_Position")`.
-- `dtWfoll: 0.22 (4446.9hz)` which is the number of milliseconds it took to set a new goal position for the follower arm using `follower_arm.write("Goal_position", leader_pos)` ; note that writing is done asynchronously so it takes less time than reading.
-
-Importantly: If you don't have any camera, you can remove them dynamically with this [draccus](https://github.com/dlwh/draccus) syntax `--robot.cameras='{}'`:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=koch \
-  --robot.cameras='{}' \
-  --control.type=teleoperate
-```
-
-We advise to create a new yaml file when the command becomes too long.
-
-## 3. Record your Dataset and Visualize it
-
-Using what you've learned previously, you can now easily record a dataset of states and actions for one episode. You can use `busy_wait` to control the speed of teleoperation and record at a fixed `fps` (frame per seconds).
-
-Try this code to record 30 seconds at 60 fps:
-```python
-import time
-from lerobot.scripts.control_robot import busy_wait
-
-record_time_s = 30
-fps = 60
-
-states = []
-actions = []
-for _ in range(record_time_s * fps):
-    start_time = time.perf_counter()
-    observation, action = robot.teleop_step(record_data=True)
-
-    states.append(observation["observation.state"])
-    actions.append(action["action"])
-
-    dt_s = time.perf_counter() - start_time
-    busy_wait(1 / fps - dt_s)
-
-# Note that observation and action are available in RAM, but
-# you could potentially store them on disk with pickle/hdf5 or
-# our optimized format `LeRobotDataset`. More on this next.
-```
-
-Importantly, many utilities are still missing. For instance, if you have cameras, you will need to save the images on disk to not go out of RAM, and to do so in threads to not slow down communication with your robot. Also, you will need to store your data in a format optimized for training and web sharing like [`LeRobotDataset`](../lerobot/common/datasets/lerobot_dataset.py). More on this in the next section.
-
-### a. Use the `record` function
-
-You can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) to achieve efficient data recording. It encompasses many recording utilities:
-1. Frames from cameras are saved on disk in threads, and encoded into videos at the end of each episode recording.
-2. Video streams from cameras are displayed in window so that you can verify them.
-3. Data is stored with [`LeRobotDataset`](../lerobot/common/datasets/lerobot_dataset.py) format which is pushed to your Hugging Face page (unless `--control.push_to_hub=false` is provided).
-4. Checkpoints are done during recording, so if any issue occurs, you can resume recording by re-running the same command again with `--control.resume=true`. You will need to manually delete the dataset directory if you want to start recording from scratch.
-5. Set the flow of data recording using command line arguments:
-   - `--control.warmup_time_s=10` defines the number of seconds before starting data collection. It allows the robot devices to warmup and synchronize (10 seconds by default).
-   - `--control.episode_time_s=60` defines the number of seconds for data recording for each episode (60 seconds by default).
-   - `--control.reset_time_s=60` defines the number of seconds for resetting the environment after each episode (60 seconds by default).
-   - `--control.num_episodes=50` defines the number of episodes to record (50 by default).
-6. Control the flow during data recording using keyboard keys:
-   - Press right arrow `->` at any time during episode recording to early stop and go to resetting. Same during resetting, to early stop and to go to the next episode recording.
-   - Press left arrow `<-` at any time during episode recording or resetting to early stop, cancel the current episode, and re-record it.
-   - Press escape `ESC` at any time during episode recording to end the session early and go straight to video encoding and dataset uploading.
-7. Similarly to `teleoperate`, you can also use the command line to override anything.
-
-Before trying `record`, if you want to push your dataset to the hub, make sure you've logged in using a write-access token, which can be generated from the [Hugging Face settings](https://huggingface.co/settings/tokens):
-```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
-```
-Also, store your Hugging Face repository name in a variable (e.g. `cadene` or `lerobot`). For instance, run this to use your Hugging Face user name as repository:
-```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
-echo $HF_USER
-```
-If you don't want to push to hub, use `--control.push_to_hub=false`.
-
-Now run this to record 2 episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=koch \
-  --control.type=record \
-  --control.single_task="Grasp a lego block and put it in the bin." \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/koch_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=2 \
-  --control.push_to_hub=true
-```
-
-
-This will write your dataset locally to `~/.cache/huggingface/lerobot/{repo-id}` (e.g. `data/cadene/koch_test`) and push it on the hub at `https://huggingface.co/datasets/{HF_USER}/{repo-id}`. Your dataset will be automatically tagged with `LeRobot` for the community to find it easily, and you can also add custom tags (in this case `tutorial` for example).
-
-You can look for other LeRobot datasets on the hub by searching for `LeRobot` tags: https://huggingface.co/datasets?other=LeRobot
-
-You will see a lot of lines appearing like this one:
-```
-INFO 2024-08-10 15:02:58 ol_robot.py:219 dt:33.34 (30.0hz) dtRlead: 5.06 (197.5hz) dtWfoll: 0.25 (3963.7hz) dtRfoll: 6.22 (160.7hz) dtRlaptop: 32.57 (30.7hz) dtRphone: 33.84 (29.5hz)
-```
-It contains:
-- `2024-08-10 15:02:58` which is the date and time of the call to the print function,
-- `ol_robot.py:219` which is the end of the file name and the line number where the print function is called  (`lerobot/scripts/control_robot.py` line `219`).
-- `dt:33.34 (30.0hz)` which is the "delta time" or the number of milliseconds spent between the previous call to `robot.teleop_step(record_data=True)` and the current one, associated with the frequency (33.34 ms equals 30.0 Hz) ; note that we use `--fps 30` so we expect 30.0 Hz ; when a step takes more time, the line appears in yellow.
-- `dtRlead: 5.06 (197.5hz)` which is the delta time of reading the present position of the leader arm.
-- `dtWfoll: 0.25 (3963.7hz)` which is the delta time of writing the goal position on the follower arm ; writing is asynchronous so it takes less time than reading.
-- `dtRfoll: 6.22 (160.7hz)` which is the delta time of reading the present position on the follower arm.
-- `dtRlaptop:32.57 (30.7hz) ` which is the delta time of capturing an image from the laptop camera in the thread running asynchronously.
-- `dtRphone:33.84 (29.5hz)` which is the delta time of capturing an image from the phone camera in the thread running asynchronously.
-
-Troubleshooting:
-- On Linux, if the left and right arrow keys and escape key don't have any effect during data recording, make sure you've set the `$DISPLAY` environment variable. See [pynput limitations](https://pynput.readthedocs.io/en/latest/limitations.html#linux).
-
-At the end of data recording, your dataset will be uploaded on your Hugging Face page (e.g. https://huggingface.co/datasets/cadene/koch_test) that you can obtain by running:
-```bash
-echo https://huggingface.co/datasets/${HF_USER}/koch_test
-```
-
-### b. Advice for recording dataset
-
-Once you're comfortable with data recording, it's time to create a larger dataset for training. A good starting task is grasping an object at different locations and placing it in a bin. We suggest recording at least 50 episodes, with 10 episodes per location. Keep the cameras fixed and maintain consistent grasping behavior throughout the recordings.
-
-In the following sections, you’ll train your neural network. After achieving reliable grasping performance, you can start introducing more variations during data collection, such as additional grasp locations, different grasping techniques, and altering camera positions.
-
-Avoid adding too much variation too quickly, as it may hinder your results.
-
-In the coming months, we plan to release a foundational model for robotics. We anticipate that fine-tuning this model will enhance generalization, reducing the need for strict consistency during data collection.
-
-### c. Visualize all episodes
-
-You can visualize your dataset by running:
-```bash
-python lerobot/scripts/visualize_dataset_html.py \
-  --repo-id ${HF_USER}/koch_test
-```
-
-Note: You might need to add `--local-files-only 1` if your dataset was not uploaded to hugging face hub.
-
-This will launch a local web server that looks like this:
-<div style="text-align:center;">
-  <img src="../media/tutorial/visualize_dataset_html.webp?raw=true" alt="Koch v1.1 leader and follower arms" title="Koch v1.1 leader and follower arms" width="100%">
-</div>
-
-### d. Replay episode on your robot with the `replay` function
-
-A useful feature of [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) is the `replay` function, which allows to replay on your robot any episode that you've recorded or episodes from any dataset out there. This function helps you test the repeatability of your robot's actions and assess transferability across robots of the same model.
-
-To replay the first episode of the dataset you just recorded, run the following command:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=koch \
-  --control.type=replay \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/koch_test \
-  --control.episode=0
-```
-
-Your robot should replicate movements similar to those you recorded. For example, check out [this video](https://x.com/RemiCadene/status/1793654950905680090) where we use `replay` on a Aloha robot from [Trossen Robotics](https://www.trossenrobotics.com).
-
-## 4. Train a policy on your data
-
-### a. Use the `train` script
-
-To train a policy to control your robot, use the [`python lerobot/scripts/train.py`](../lerobot/scripts/train.py) script. A few arguments are required. Here is an example command:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/koch_test \
-  --policy.type=act \
-  --output_dir=outputs/train/act_koch_test \
-  --job_name=act_koch_test \
-  --policy.device=cuda \
-  --wandb.enable=true
-```
-
-Let's explain it:
-1. We provided the dataset as argument with `--dataset.repo_id=${HF_USER}/koch_test`.
-2. We provided the policy with `policy.type=act`. This loads configurations from [`configuration_act.py`](../lerobot/common/policies/act/configuration_act.py). Importantly, this policy will automatically adapt to the number of motor sates, motor actions and cameras of your robot (e.g. `laptop` and `phone`) which have been saved in your dataset.
-4. We provided `policy.device=cuda` since we are training on a Nvidia GPU, but you could use `policy.device=mps` to train on Apple silicon.
-5. We provided `wandb.enable=true` to use [Weights and Biases](https://docs.wandb.ai/quickstart) for visualizing training plots. This is optional but if you use it, make sure you are logged in by running `wandb login`.
-
-For more information on the `train` script see the previous tutorial: [`examples/4_train_policy_with_script.md`](../examples/4_train_policy_with_script.md)
-
-### b. (Optional) Upload policy checkpoints to the hub
-
-Once training is done, upload the latest checkpoint with:
-```bash
-huggingface-cli upload ${HF_USER}/act_koch_test \
-  outputs/train/act_koch_test/checkpoints/last/pretrained_model
-```
-
-You can also upload intermediate checkpoints with:
-```bash
-CKPT=010000
-huggingface-cli upload ${HF_USER}/act_koch_test_${CKPT} \
-  outputs/train/act_koch_test/checkpoints/${CKPT}/pretrained_model
-```
-
-## 5. Evaluate your policy
-
-Now that you have a policy checkpoint, you can easily control your robot with it using methods from [`ManipulatorRobot`](../lerobot/common/robot_devices/robots/manipulator.py) and the policy.
-
-Try this code for running inference for 60 seconds at 30 fps:
-```python
-from lerobot.common.policies.act.modeling_act import ACTPolicy
-
-inference_time_s = 60
-fps = 30
-device = "cuda"  # TODO: On Mac, use "mps" or "cpu"
-
-ckpt_path = "outputs/train/act_koch_test/checkpoints/last/pretrained_model"
-policy = ACTPolicy.from_pretrained(ckpt_path)
-policy.to(device)
-
-for _ in range(inference_time_s * fps):
-    start_time = time.perf_counter()
-
-    # Read the follower state and access the frames from the cameras
-    observation = robot.capture_observation()
-
-    # Convert to pytorch format: channel first and float32 in [0,1]
-    # with batch dimension
-    for name in observation:
-        if "image" in name:
-            observation[name] = observation[name].type(torch.float32) / 255
-            observation[name] = observation[name].permute(2, 0, 1).contiguous()
-        observation[name] = observation[name].unsqueeze(0)
-        observation[name] = observation[name].to(device)
-
-    # Compute the next action with the policy
-    # based on the current observation
-    action = policy.select_action(observation)
-    # Remove batch dimension
-    action = action.squeeze(0)
-    # Move to cpu, if not already the case
-    action = action.to("cpu")
-    # Order the robot to move
-    robot.send_action(action)
-
-    dt_s = time.perf_counter() - start_time
-    busy_wait(1 / fps - dt_s)
-```
-
-### a. Use our `record` function
-
-Ideally, when controlling your robot with your neural network, you would want to record evaluation episodes and to be able to visualize them later on, or even train on them like in Reinforcement Learning. This pretty much corresponds to recording a new dataset but with a neural network providing the actions instead of teleoperation.
-
-To this end, you can use the `record` function from [`lerobot/scripts/control_robot.py`](../lerobot/scripts/control_robot.py) but with a policy checkpoint as input. For instance, run this command to record 10 evaluation episodes:
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=koch \
-  --control.type=record \
-  --control.fps=30 \
-  --control.repo_id=${HF_USER}/eval_act_koch_test \
-  --control.tags='["tutorial"]' \
-  --control.warmup_time_s=5 \
-  --control.episode_time_s=30 \
-  --control.reset_time_s=30 \
-  --control.num_episodes=10 \
-  --control.push_to_hub=true \
-  --control.policy.path=outputs/train/act_koch_test/checkpoints/last/pretrained_model
-```
-
-As you can see, it's almost the same command as previously used to record your training dataset. Two things changed:
-1. There is an additional `--control.policy.path` argument which indicates the path to your policy checkpoint with  (e.g. `outputs/train/eval_koch_test/checkpoints/last/pretrained_model`). You can also use the model repository if you uploaded a model checkpoint to the hub (e.g. `${HF_USER}/act_koch_test`).
-2. The name of dataset begins by `eval` to reflect that you are running inference (e.g. `${HF_USER}/eval_act_koch_test`).
-
-### b. Visualize evaluation afterwards
-
-You can then visualize your evaluation dataset by running the same command as before but with the new inference dataset as argument:
-```bash
-python lerobot/scripts/visualize_dataset.py \
-  --repo-id ${HF_USER}/eval_act_koch_test
-```
-
-## 6. Next step
-
-Join our [Discord](https://discord.com/invite/s3KuuzsPFb) to collaborate on data collection and help us train a fully open-source foundational models for robotics!

examples/advanced/1_add_image_transforms.py

@@ -22,7 +22,7 @@ from pathlib import Path
 
 from torchvision.transforms import ToPILImage, v2
 
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 
 dataset_repo_id = "lerobot/aloha_static_screw_driver"
 

examples/advanced/2_calculate_validation_loss.py

@@ -26,8 +26,8 @@ import math
 
 import torch
 
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
-from lerobot.common.policies.diffusion.modeling_diffusion import DiffusionPolicy
+from lerobot.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
+from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
 
 
 def main():

examples/backward_compatibility/replay.py

@@ -0,0 +1,105 @@
+# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Replays the actions of an episode from a dataset on a robot.
+
+Example:
+
+```shell
+python -m lerobot.replay \
+    --robot.type=so100_follower \
+    --robot.port=/dev/tty.usbmodem58760431541 \
+    --robot.id=black \
+    --dataset.repo_id=aliberts/record-test \
+    --dataset.episode=2
+```
+"""
+
+import logging
+import time
+from dataclasses import asdict, dataclass
+from pathlib import Path
+from pprint import pformat
+
+import draccus
+
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.robots import (  # noqa: F401
+    Robot,
+    RobotConfig,
+    koch_follower,
+    make_robot_from_config,
+    so100_follower,
+    so101_follower,
+)
+from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.utils import (
+    init_logging,
+    log_say,
+)
+
+
+@dataclass
+class DatasetReplayConfig:
+    # Dataset identifier. By convention it should match '{hf_username}/{dataset_name}' (e.g. `lerobot/test`).
+    repo_id: str
+    # Episode to replay.
+    episode: int
+    # Root directory where the dataset will be stored (e.g. 'dataset/path').
+    root: str | Path | None = None
+    # Limit the frames per second. By default, uses the policy fps.
+    fps: int = 30
+
+
+@dataclass
+class ReplayConfig:
+    robot: RobotConfig
+    dataset: DatasetReplayConfig
+    # Use vocal synthesis to read events.
+    play_sounds: bool = True
+
+
+@draccus.wrap()
+def replay(cfg: ReplayConfig):
+    init_logging()
+    logging.info(pformat(asdict(cfg)))
+
+    robot = make_robot_from_config(cfg.robot)
+    dataset = LeRobotDataset(cfg.dataset.repo_id, root=cfg.dataset.root, episodes=[cfg.dataset.episode])
+    actions = dataset.hf_dataset.select_columns("action")
+    robot.connect()
+
+    log_say("Replaying episode", cfg.play_sounds, blocking=True)
+    for idx in range(dataset.num_frames):
+        start_episode_t = time.perf_counter()
+
+        action_array = actions[idx]["action"]
+        action = {}
+        for i, name in enumerate(dataset.features["action"]["names"]):
+            key = f"{name.removeprefix('main_')}.pos"
+            action[key] = action_array[i].item()
+
+        action["shoulder_lift.pos"] = -(action["shoulder_lift.pos"] - 90)
+        action["elbow_flex.pos"] -= 90
+        robot.send_action(action)
+
+        dt_s = time.perf_counter() - start_episode_t
+        busy_wait(1 / dataset.fps - dt_s)
+
+    robot.disconnect()
+
+
+if __name__ == "__main__":
+    replay()

examples/lekiwi/evaluate.py

@@ -0,0 +1,90 @@
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.utils import hw_to_dataset_features
+from lerobot.policies.act.modeling_act import ACTPolicy
+from lerobot.record import record_loop
+from lerobot.robots.lekiwi import LeKiwiClient, LeKiwiClientConfig
+from lerobot.utils.control_utils import init_keyboard_listener
+from lerobot.utils.utils import log_say
+from lerobot.utils.visualization_utils import _init_rerun
+
+NUM_EPISODES = 2
+FPS = 30
+EPISODE_TIME_SEC = 60
+TASK_DESCRIPTION = "My task description"
+
+# Create the robot and teleoperator configurations
+robot_config = LeKiwiClientConfig(remote_ip="172.18.134.136", id="lekiwi")
+robot = LeKiwiClient(robot_config)
+
+policy = ACTPolicy.from_pretrained("<hf_username>/<policy_repo_id>")
+
+# Configure the dataset features
+action_features = hw_to_dataset_features(robot.action_features, "action")
+obs_features = hw_to_dataset_features(robot.observation_features, "observation")
+dataset_features = {**action_features, **obs_features}
+
+# Create the dataset
+dataset = LeRobotDataset.create(
+    repo_id="<hf_username>/<eval_dataset_repo_id>",
+    fps=FPS,
+    features=dataset_features,
+    robot_type=robot.name,
+    use_videos=True,
+    image_writer_threads=4,
+)
+
+# To connect you already should have this script running on LeKiwi: `python -m lerobot.robots.lekiwi.lekiwi_host --robot.id=my_awesome_kiwi`
+robot.connect()
+
+_init_rerun(session_name="recording")
+
+listener, events = init_keyboard_listener()
+
+if not robot.is_connected:
+    raise ValueError("Robot is not connected!")
+
+recorded_episodes = 0
+while recorded_episodes < NUM_EPISODES and not events["stop_recording"]:
+    log_say(f"Running inference, recording eval episode {recorded_episodes} of {NUM_EPISODES}")
+
+    # Run the policy inference loop
+    record_loop(
+        robot=robot,
+        events=events,
+        fps=FPS,
+        policy=policy,
+        dataset=dataset,
+        control_time_s=EPISODE_TIME_SEC,
+        single_task=TASK_DESCRIPTION,
+        display_data=True,
+    )
+
+    # Logic for reset env
+    if not events["stop_recording"] and (
+        (recorded_episodes < NUM_EPISODES - 1) or events["rerecord_episode"]
+    ):
+        log_say("Reset the environment")
+        record_loop(
+            robot=robot,
+            events=events,
+            fps=FPS,
+            control_time_s=EPISODE_TIME_SEC,
+            single_task=TASK_DESCRIPTION,
+            display_data=True,
+        )
+
+    if events["rerecord_episode"]:
+        log_say("Re-record episode")
+        events["rerecord_episode"] = False
+        events["exit_early"] = False
+        dataset.clear_episode_buffer()
+        continue
+
+    dataset.save_episode()
+    recorded_episodes += 1
+
+# Upload to hub and clean up
+dataset.push_to_hub()
+
+robot.disconnect()
+listener.stop()

examples/lekiwi/record.py

@@ -0,0 +1,101 @@
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.utils import hw_to_dataset_features
+from lerobot.record import record_loop
+from lerobot.robots.lekiwi.config_lekiwi import LeKiwiClientConfig
+from lerobot.robots.lekiwi.lekiwi_client import LeKiwiClient
+from lerobot.teleoperators.keyboard import KeyboardTeleop, KeyboardTeleopConfig
+from lerobot.teleoperators.so100_leader import SO100Leader, SO100LeaderConfig
+from lerobot.utils.control_utils import init_keyboard_listener
+from lerobot.utils.utils import log_say
+from lerobot.utils.visualization_utils import _init_rerun
+
+NUM_EPISODES = 3
+FPS = 30
+EPISODE_TIME_SEC = 30
+RESET_TIME_SEC = 10
+TASK_DESCRIPTION = "My task description"
+
+# Create the robot and teleoperator configurations
+robot_config = LeKiwiClientConfig(remote_ip="172.18.134.136", id="lekiwi")
+leader_arm_config = SO100LeaderConfig(port="/dev/tty.usbmodem585A0077581", id="my_awesome_leader_arm")
+keyboard_config = KeyboardTeleopConfig()
+
+robot = LeKiwiClient(robot_config)
+leader_arm = SO100Leader(leader_arm_config)
+keyboard = KeyboardTeleop(keyboard_config)
+
+# Configure the dataset features
+action_features = hw_to_dataset_features(robot.action_features, "action")
+obs_features = hw_to_dataset_features(robot.observation_features, "observation")
+dataset_features = {**action_features, **obs_features}
+
+# Create the dataset
+dataset = LeRobotDataset.create(
+    repo_id="<hf_username>/<dataset_repo_id>",
+    fps=FPS,
+    features=dataset_features,
+    robot_type=robot.name,
+    use_videos=True,
+    image_writer_threads=4,
+)
+
+# To connect you already should have this script running on LeKiwi: `python -m lerobot.robots.lekiwi.lekiwi_host --robot.id=my_awesome_kiwi`
+robot.connect()
+leader_arm.connect()
+keyboard.connect()
+
+_init_rerun(session_name="lekiwi_record")
+
+listener, events = init_keyboard_listener()
+
+if not robot.is_connected or not leader_arm.is_connected or not keyboard.is_connected:
+    raise ValueError("Robot, leader arm of keyboard is not connected!")
+
+recorded_episodes = 0
+while recorded_episodes < NUM_EPISODES and not events["stop_recording"]:
+    log_say(f"Recording episode {recorded_episodes}")
+
+    # Run the record loop
+    record_loop(
+        robot=robot,
+        events=events,
+        fps=FPS,
+        dataset=dataset,
+        teleop=[leader_arm, keyboard],
+        control_time_s=EPISODE_TIME_SEC,
+        single_task=TASK_DESCRIPTION,
+        display_data=True,
+    )
+
+    # Logic for reset env
+    if not events["stop_recording"] and (
+        (recorded_episodes < NUM_EPISODES - 1) or events["rerecord_episode"]
+    ):
+        log_say("Reset the environment")
+        record_loop(
+            robot=robot,
+            events=events,
+            fps=FPS,
+            teleop=[leader_arm, keyboard],
+            control_time_s=RESET_TIME_SEC,
+            single_task=TASK_DESCRIPTION,
+            display_data=True,
+        )
+
+    if events["rerecord_episode"]:
+        log_say("Re-record episode")
+        events["rerecord_episode"] = False
+        events["exit_early"] = False
+        dataset.clear_episode_buffer()
+        continue
+
+    dataset.save_episode()
+    recorded_episodes += 1
+
+# Upload to hub and clean up
+dataset.push_to_hub()
+
+robot.disconnect()
+leader_arm.disconnect()
+keyboard.disconnect()
+listener.stop()

examples/lekiwi/replay.py

@@ -0,0 +1,33 @@
+import time
+
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.robots.lekiwi.config_lekiwi import LeKiwiClientConfig
+from lerobot.robots.lekiwi.lekiwi_client import LeKiwiClient
+from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.utils import log_say
+
+EPISODE_IDX = 0
+
+robot_config = LeKiwiClientConfig(remote_ip="172.18.134.136", id="lekiwi")
+robot = LeKiwiClient(robot_config)
+
+dataset = LeRobotDataset("<hf_username>/<dataset_repo_id>", episodes=[EPISODE_IDX])
+actions = dataset.hf_dataset.select_columns("action")
+
+robot.connect()
+
+if not robot.is_connected:
+    raise ValueError("Robot is not connected!")
+
+log_say(f"Replaying episode {EPISODE_IDX}")
+for idx in range(dataset.num_frames):
+    t0 = time.perf_counter()
+
+    action = {
+        name: float(actions[idx]["action"][i]) for i, name in enumerate(dataset.features["action"]["names"])
+    }
+    robot.send_action(action)
+
+    busy_wait(max(1.0 / dataset.fps - (time.perf_counter() - t0), 0.0))
+
+robot.disconnect()

examples/lekiwi/teleoperate.py

@@ -0,0 +1,47 @@
+import time
+
+from lerobot.robots.lekiwi import LeKiwiClient, LeKiwiClientConfig
+from lerobot.teleoperators.keyboard.teleop_keyboard import KeyboardTeleop, KeyboardTeleopConfig
+from lerobot.teleoperators.so100_leader import SO100Leader, SO100LeaderConfig
+from lerobot.utils.robot_utils import busy_wait
+from lerobot.utils.visualization_utils import _init_rerun, log_rerun_data
+
+FPS = 30
+
+# Create the robot and teleoperator configurations
+robot_config = LeKiwiClientConfig(remote_ip="172.18.134.136", id="my_lekiwi")
+teleop_arm_config = SO100LeaderConfig(port="/dev/tty.usbmodem585A0077581", id="my_awesome_leader_arm")
+keyboard_config = KeyboardTeleopConfig(id="my_laptop_keyboard")
+
+robot = LeKiwiClient(robot_config)
+leader_arm = SO100Leader(teleop_arm_config)
+keyboard = KeyboardTeleop(keyboard_config)
+
+# To connect you already should have this script running on LeKiwi: `python -m lerobot.robots.lekiwi.lekiwi_host --robot.id=my_awesome_kiwi`
+robot.connect()
+leader_arm.connect()
+keyboard.connect()
+
+_init_rerun(session_name="lekiwi_teleop")
+
+if not robot.is_connected or not leader_arm.is_connected or not keyboard.is_connected:
+    raise ValueError("Robot, leader arm of keyboard is not connected!")
+
+while True:
+    t0 = time.perf_counter()
+
+    observation = robot.get_observation()
+
+    arm_action = leader_arm.get_action()
+    arm_action = {f"arm_{k}": v for k, v in arm_action.items()}
+
+    keyboard_keys = keyboard.get_action()
+    base_action = robot._from_keyboard_to_base_action(keyboard_keys)
+
+    log_rerun_data(observation, {**arm_action, **base_action})
+
+    action = {**arm_action, **base_action} if len(base_action) > 0 else arm_action
+
+    robot.send_action(action)
+
+    busy_wait(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))

lerobot/common/optim/optimizers.py

@@ -1,118 +0,0 @@
-#!/usr/bin/env python
-
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-import abc
-from dataclasses import asdict, dataclass
-from pathlib import Path
-
-import draccus
-import torch
-from safetensors.torch import load_file, save_file
-
-from lerobot.common.constants import (
-    OPTIMIZER_PARAM_GROUPS,
-    OPTIMIZER_STATE,
-)
-from lerobot.common.datasets.utils import flatten_dict, unflatten_dict, write_json
-from lerobot.common.utils.io_utils import deserialize_json_into_object
-
-
-@dataclass
-class OptimizerConfig(draccus.ChoiceRegistry, abc.ABC):
-    lr: float
-    weight_decay: float
-    grad_clip_norm: float
-
-    @property
-    def type(self) -> str:
-        return self.get_choice_name(self.__class__)
-
-    @classmethod
-    def default_choice_name(cls) -> str | None:
-        return "adam"
-
-    @abc.abstractmethod
-    def build(self) -> torch.optim.Optimizer:
-        raise NotImplementedError
-
-
-@OptimizerConfig.register_subclass("adam")
-@dataclass
-class AdamConfig(OptimizerConfig):
-    lr: float = 1e-3
-    betas: tuple[float, float] = (0.9, 0.999)
-    eps: float = 1e-8
-    weight_decay: float = 0.0
-    grad_clip_norm: float = 10.0
-
-    def build(self, params: dict) -> torch.optim.Optimizer:
-        kwargs = asdict(self)
-        kwargs.pop("grad_clip_norm")
-        return torch.optim.Adam(params, **kwargs)
-
-
-@OptimizerConfig.register_subclass("adamw")
-@dataclass
-class AdamWConfig(OptimizerConfig):
-    lr: float = 1e-3
-    betas: tuple[float, float] = (0.9, 0.999)
-    eps: float = 1e-8
-    weight_decay: float = 1e-2
-    grad_clip_norm: float = 10.0
-
-    def build(self, params: dict) -> torch.optim.Optimizer:
-        kwargs = asdict(self)
-        kwargs.pop("grad_clip_norm")
-        return torch.optim.AdamW(params, **kwargs)
-
-
-@OptimizerConfig.register_subclass("sgd")
-@dataclass
-class SGDConfig(OptimizerConfig):
-    lr: float = 1e-3
-    momentum: float = 0.0
-    dampening: float = 0.0
-    nesterov: bool = False
-    weight_decay: float = 0.0
-    grad_clip_norm: float = 10.0
-
-    def build(self, params: dict) -> torch.optim.Optimizer:
-        kwargs = asdict(self)
-        kwargs.pop("grad_clip_norm")
-        return torch.optim.SGD(params, **kwargs)
-
-
-def save_optimizer_state(optimizer: torch.optim.Optimizer, save_dir: Path) -> None:
-    state = optimizer.state_dict()
-    param_groups = state.pop("param_groups")
-    flat_state = flatten_dict(state)
-    save_file(flat_state, save_dir / OPTIMIZER_STATE)
-    write_json(param_groups, save_dir / OPTIMIZER_PARAM_GROUPS)
-
-
-def load_optimizer_state(optimizer: torch.optim.Optimizer, save_dir: Path) -> torch.optim.Optimizer:
-    current_state_dict = optimizer.state_dict()
-    flat_state = load_file(save_dir / OPTIMIZER_STATE)
-    state = unflatten_dict(flat_state)
-    loaded_state_dict = {"state": {int(k): v for k, v in state["state"].items()}}
-
-    if "param_groups" in current_state_dict:
-        param_groups = deserialize_json_into_object(
-            save_dir / OPTIMIZER_PARAM_GROUPS, current_state_dict["param_groups"]
-        )
-        loaded_state_dict["param_groups"] = param_groups
-
-    optimizer.load_state_dict(loaded_state_dict)
-    return optimizer

lerobot/common/robot_devices/cameras/configs.py

@@ -1,114 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import abc
-from dataclasses import dataclass
-
-import draccus
-
-
-@dataclass
-class CameraConfig(draccus.ChoiceRegistry, abc.ABC):
-    @property
-    def type(self) -> str:
-        return self.get_choice_name(self.__class__)
-
-
-@CameraConfig.register_subclass("opencv")
-@dataclass
-class OpenCVCameraConfig(CameraConfig):
-    """
-    Example of tested options for Intel Real Sense D405:
-
-    ```python
-    OpenCVCameraConfig(0, 30, 640, 480)
-    OpenCVCameraConfig(0, 60, 640, 480)
-    OpenCVCameraConfig(0, 90, 640, 480)
-    OpenCVCameraConfig(0, 30, 1280, 720)
-    ```
-    """
-
-    camera_index: int
-    fps: int | None = None
-    width: int | None = None
-    height: int | None = None
-    color_mode: str = "rgb"
-    channels: int | None = None
-    rotation: int | None = None
-    mock: bool = False
-
-    def __post_init__(self):
-        if self.color_mode not in ["rgb", "bgr"]:
-            raise ValueError(
-                f"`color_mode` is expected to be 'rgb' or 'bgr', but {self.color_mode} is provided."
-            )
-
-        self.channels = 3
-
-        if self.rotation not in [-90, None, 90, 180]:
-            raise ValueError(f"`rotation` must be in [-90, None, 90, 180] (got {self.rotation})")
-
-
-@CameraConfig.register_subclass("intelrealsense")
-@dataclass
-class IntelRealSenseCameraConfig(CameraConfig):
-    """
-    Example of tested options for Intel Real Sense D405:
-
-    ```python
-    IntelRealSenseCameraConfig(128422271347, 30, 640, 480)
-    IntelRealSenseCameraConfig(128422271347, 60, 640, 480)
-    IntelRealSenseCameraConfig(128422271347, 90, 640, 480)
-    IntelRealSenseCameraConfig(128422271347, 30, 1280, 720)
-    IntelRealSenseCameraConfig(128422271347, 30, 640, 480, use_depth=True)
-    IntelRealSenseCameraConfig(128422271347, 30, 640, 480, rotation=90)
-    ```
-    """
-
-    name: str | None = None
-    serial_number: int | None = None
-    fps: int | None = None
-    width: int | None = None
-    height: int | None = None
-    color_mode: str = "rgb"
-    channels: int | None = None
-    use_depth: bool = False
-    force_hardware_reset: bool = True
-    rotation: int | None = None
-    mock: bool = False
-
-    def __post_init__(self):
-        # bool is stronger than is None, since it works with empty strings
-        if bool(self.name) and bool(self.serial_number):
-            raise ValueError(
-                f"One of them must be set: name or serial_number, but {self.name=} and {self.serial_number=} provided."
-            )
-
-        if self.color_mode not in ["rgb", "bgr"]:
-            raise ValueError(
-                f"`color_mode` is expected to be 'rgb' or 'bgr', but {self.color_mode} is provided."
-            )
-
-        self.channels = 3
-
-        at_least_one_is_not_none = self.fps is not None or self.width is not None or self.height is not None
-        at_least_one_is_none = self.fps is None or self.width is None or self.height is None
-        if at_least_one_is_not_none and at_least_one_is_none:
-            raise ValueError(
-                "For `fps`, `width` and `height`, either all of them need to be set, or none of them, "
-                f"but {self.fps=}, {self.width=}, {self.height=} were provided."
-            )
-
-        if self.rotation not in [-90, None, 90, 180]:
-            raise ValueError(f"`rotation` must be in [-90, None, 90, 180] (got {self.rotation})")

lerobot/common/robot_devices/cameras/intelrealsense.py

@@ -1,538 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-"""
-This file contains utilities for recording frames from Intel Realsense cameras.
-"""
-
-import argparse
-import concurrent.futures
-import logging
-import math
-import shutil
-import threading
-import time
-import traceback
-from collections import Counter
-from pathlib import Path
-from threading import Thread
-
-import numpy as np
-from PIL import Image
-
-from lerobot.common.robot_devices.cameras.configs import IntelRealSenseCameraConfig
-from lerobot.common.robot_devices.utils import (
-    RobotDeviceAlreadyConnectedError,
-    RobotDeviceNotConnectedError,
-    busy_wait,
-)
-from lerobot.common.utils.utils import capture_timestamp_utc
-
-SERIAL_NUMBER_INDEX = 1
-
-
-def find_cameras(raise_when_empty=True, mock=False) -> list[dict]:
-    """
-    Find the names and the serial numbers of the Intel RealSense cameras
-    connected to the computer.
-    """
-    if mock:
-        import tests.cameras.mock_pyrealsense2 as rs
-    else:
-        import pyrealsense2 as rs
-
-    cameras = []
-    for device in rs.context().query_devices():
-        serial_number = int(device.get_info(rs.camera_info(SERIAL_NUMBER_INDEX)))
-        name = device.get_info(rs.camera_info.name)
-        cameras.append(
-            {
-                "serial_number": serial_number,
-                "name": name,
-            }
-        )
-
-    if raise_when_empty and len(cameras) == 0:
-        raise OSError(
-            "Not a single camera was detected. Try re-plugging, or re-installing `librealsense` and its python wrapper `pyrealsense2`, or updating the firmware."
-        )
-
-    return cameras
-
-
-def save_image(img_array, serial_number, frame_index, images_dir):
-    try:
-        img = Image.fromarray(img_array)
-        path = images_dir / f"camera_{serial_number}_frame_{frame_index:06d}.png"
-        path.parent.mkdir(parents=True, exist_ok=True)
-        img.save(str(path), quality=100)
-        logging.info(f"Saved image: {path}")
-    except Exception as e:
-        logging.error(f"Failed to save image for camera {serial_number} frame {frame_index}: {e}")
-
-
-def save_images_from_cameras(
-    images_dir: Path,
-    serial_numbers: list[int] | None = None,
-    fps=None,
-    width=None,
-    height=None,
-    record_time_s=2,
-    mock=False,
-):
-    """
-    Initializes all the cameras and saves images to the directory. Useful to visually identify the camera
-    associated to a given serial number.
-    """
-    if serial_numbers is None or len(serial_numbers) == 0:
-        camera_infos = find_cameras(mock=mock)
-        serial_numbers = [cam["serial_number"] for cam in camera_infos]
-
-    if mock:
-        import tests.cameras.mock_cv2 as cv2
-    else:
-        import cv2
-
-    print("Connecting cameras")
-    cameras = []
-    for cam_sn in serial_numbers:
-        print(f"{cam_sn=}")
-        config = IntelRealSenseCameraConfig(
-            serial_number=cam_sn, fps=fps, width=width, height=height, mock=mock
-        )
-        camera = IntelRealSenseCamera(config)
-        camera.connect()
-        print(
-            f"IntelRealSenseCamera({camera.serial_number}, fps={camera.fps}, width={camera.capture_width}, height={camera.capture_height}, color_mode={camera.color_mode})"
-        )
-        cameras.append(camera)
-
-    images_dir = Path(images_dir)
-    if images_dir.exists():
-        shutil.rmtree(
-            images_dir,
-        )
-    images_dir.mkdir(parents=True, exist_ok=True)
-
-    print(f"Saving images to {images_dir}")
-    frame_index = 0
-    start_time = time.perf_counter()
-    try:
-        with concurrent.futures.ThreadPoolExecutor(max_workers=1) as executor:
-            while True:
-                now = time.perf_counter()
-
-                for camera in cameras:
-                    # If we use async_read when fps is None, the loop will go full speed, and we will end up
-                    # saving the same images from the cameras multiple times until the RAM/disk is full.
-                    image = camera.read() if fps is None else camera.async_read()
-                    if image is None:
-                        print("No Frame")
-
-                    bgr_converted_image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
-
-                    executor.submit(
-                        save_image,
-                        bgr_converted_image,
-                        camera.serial_number,
-                        frame_index,
-                        images_dir,
-                    )
-
-                if fps is not None:
-                    dt_s = time.perf_counter() - now
-                    busy_wait(1 / fps - dt_s)
-
-                if time.perf_counter() - start_time > record_time_s:
-                    break
-
-                print(f"Frame: {frame_index:04d}\tLatency (ms): {(time.perf_counter() - now) * 1000:.2f}")
-
-                frame_index += 1
-    finally:
-        print(f"Images have been saved to {images_dir}")
-        for camera in cameras:
-            camera.disconnect()
-
-
-class IntelRealSenseCamera:
-    """
-    The IntelRealSenseCamera class is similar to OpenCVCamera class but adds additional features for Intel Real Sense cameras:
-    - is instantiated with the serial number of the camera - won't randomly change as it can be the case of OpenCVCamera for Linux,
-    - can also be instantiated with the camera's name — if it's unique — using IntelRealSenseCamera.init_from_name(),
-    - depth map can be returned.
-
-    To find the camera indices of your cameras, you can run our utility script that will save a few frames for each camera:
-    ```bash
-    python lerobot/common/robot_devices/cameras/intelrealsense.py --images-dir outputs/images_from_intelrealsense_cameras
-    ```
-
-    When an IntelRealSenseCamera is instantiated, if no specific config is provided, the default fps, width, height and color_mode
-    of the given camera will be used.
-
-    Example of instantiating with a serial number:
-    ```python
-    from lerobot.common.robot_devices.cameras.configs import IntelRealSenseCameraConfig
-
-    config = IntelRealSenseCameraConfig(serial_number=128422271347)
-    camera = IntelRealSenseCamera(config)
-    camera.connect()
-    color_image = camera.read()
-    # when done using the camera, consider disconnecting
-    camera.disconnect()
-    ```
-
-    Example of instantiating with a name if it's unique:
-    ```
-    config = IntelRealSenseCameraConfig(name="Intel RealSense D405")
-    ```
-
-    Example of changing default fps, width, height and color_mode:
-    ```python
-    config = IntelRealSenseCameraConfig(serial_number=128422271347, fps=30, width=1280, height=720)
-    config = IntelRealSenseCameraConfig(serial_number=128422271347, fps=90, width=640, height=480)
-    config = IntelRealSenseCameraConfig(serial_number=128422271347, fps=90, width=640, height=480, color_mode="bgr")
-    # Note: might error out upon `camera.connect()` if these settings are not compatible with the camera
-    ```
-
-    Example of returning depth:
-    ```python
-    config = IntelRealSenseCameraConfig(serial_number=128422271347, use_depth=True)
-    camera = IntelRealSenseCamera(config)
-    camera.connect()
-    color_image, depth_map = camera.read()
-    ```
-    """
-
-    def __init__(
-        self,
-        config: IntelRealSenseCameraConfig,
-    ):
-        self.config = config
-        if config.name is not None:
-            self.serial_number = self.find_serial_number_from_name(config.name)
-        else:
-            self.serial_number = config.serial_number
-
-        # Store the raw (capture) resolution from the config.
-        self.capture_width = config.width
-        self.capture_height = config.height
-
-        # If rotated by ±90, swap width and height.
-        if config.rotation in [-90, 90]:
-            self.width = config.height
-            self.height = config.width
-        else:
-            self.width = config.width
-            self.height = config.height
-
-        self.fps = config.fps
-        self.channels = config.channels
-        self.color_mode = config.color_mode
-        self.use_depth = config.use_depth
-        self.force_hardware_reset = config.force_hardware_reset
-        self.mock = config.mock
-
-        self.camera = None
-        self.is_connected = False
-        self.thread = None
-        self.stop_event = None
-        self.color_image = None
-        self.depth_map = None
-        self.logs = {}
-
-        if self.mock:
-            import tests.cameras.mock_cv2 as cv2
-        else:
-            import cv2
-
-        self.rotation = None
-        if config.rotation == -90:
-            self.rotation = cv2.ROTATE_90_COUNTERCLOCKWISE
-        elif config.rotation == 90:
-            self.rotation = cv2.ROTATE_90_CLOCKWISE
-        elif config.rotation == 180:
-            self.rotation = cv2.ROTATE_180
-
-    def find_serial_number_from_name(self, name):
-        camera_infos = find_cameras()
-        camera_names = [cam["name"] for cam in camera_infos]
-        this_name_count = Counter(camera_names)[name]
-        if this_name_count > 1:
-            # TODO(aliberts): Test this with multiple identical cameras (Aloha)
-            raise ValueError(
-                f"Multiple {name} cameras have been detected. Please use their serial number to instantiate them."
-            )
-
-        name_to_serial_dict = {cam["name"]: cam["serial_number"] for cam in camera_infos}
-        cam_sn = name_to_serial_dict[name]
-
-        return cam_sn
-
-    def connect(self):
-        if self.is_connected:
-            raise RobotDeviceAlreadyConnectedError(
-                f"IntelRealSenseCamera({self.serial_number}) is already connected."
-            )
-
-        if self.mock:
-            import tests.cameras.mock_pyrealsense2 as rs
-        else:
-            import pyrealsense2 as rs
-
-        config = rs.config()
-        config.enable_device(str(self.serial_number))
-
-        if self.fps and self.capture_width and self.capture_height:
-            # TODO(rcadene): can we set rgb8 directly?
-            config.enable_stream(
-                rs.stream.color, self.capture_width, self.capture_height, rs.format.rgb8, self.fps
-            )
-        else:
-            config.enable_stream(rs.stream.color)
-
-        if self.use_depth:
-            if self.fps and self.capture_width and self.capture_height:
-                config.enable_stream(
-                    rs.stream.depth, self.capture_width, self.capture_height, rs.format.z16, self.fps
-                )
-            else:
-                config.enable_stream(rs.stream.depth)
-
-        self.camera = rs.pipeline()
-        try:
-            profile = self.camera.start(config)
-            is_camera_open = True
-        except RuntimeError:
-            is_camera_open = False
-            traceback.print_exc()
-
-        # If the camera doesn't work, display the camera indices corresponding to
-        # valid cameras.
-        if not is_camera_open:
-            # Verify that the provided `serial_number` is valid before printing the traceback
-            camera_infos = find_cameras()
-            serial_numbers = [cam["serial_number"] for cam in camera_infos]
-            if self.serial_number not in serial_numbers:
-                raise ValueError(
-                    f"`serial_number` is expected to be one of these available cameras {serial_numbers}, but {self.serial_number} is provided instead. "
-                    "To find the serial number you should use, run `python lerobot/common/robot_devices/cameras/intelrealsense.py`."
-                )
-
-            raise OSError(f"Can't access IntelRealSenseCamera({self.serial_number}).")
-
-        color_stream = profile.get_stream(rs.stream.color)
-        color_profile = color_stream.as_video_stream_profile()
-        actual_fps = color_profile.fps()
-        actual_width = color_profile.width()
-        actual_height = color_profile.height()
-
-        # Using `math.isclose` since actual fps can be a float (e.g. 29.9 instead of 30)
-        if self.fps is not None and not math.isclose(self.fps, actual_fps, rel_tol=1e-3):
-            # Using `OSError` since it's a broad that encompasses issues related to device communication
-            raise OSError(
-                f"Can't set {self.fps=} for IntelRealSenseCamera({self.serial_number}). Actual value is {actual_fps}."
-            )
-        if self.capture_width is not None and self.capture_width != actual_width:
-            raise OSError(
-                f"Can't set {self.capture_width=} for IntelRealSenseCamera({self.serial_number}). Actual value is {actual_width}."
-            )
-        if self.capture_height is not None and self.capture_height != actual_height:
-            raise OSError(
-                f"Can't set {self.capture_height=} for IntelRealSenseCamera({self.serial_number}). Actual value is {actual_height}."
-            )
-
-        self.fps = round(actual_fps)
-        self.capture_width = round(actual_width)
-        self.capture_height = round(actual_height)
-
-        self.is_connected = True
-
-    def read(self, temporary_color: str | None = None) -> np.ndarray | tuple[np.ndarray, np.ndarray]:
-        """Read a frame from the camera returned in the format height x width x channels (e.g. 480 x 640 x 3)
-        of type `np.uint8`, contrarily to the pytorch format which is float channel first.
-
-        When `use_depth=True`, returns a tuple `(color_image, depth_map)` with a depth map in the format
-        height x width (e.g. 480 x 640) of type np.uint16.
-
-        Note: Reading a frame is done every `camera.fps` times per second, and it is blocking.
-        If you are reading data from other sensors, we advise to use `camera.async_read()` which is non blocking version of `camera.read()`.
-        """
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"IntelRealSenseCamera({self.serial_number}) is not connected. Try running `camera.connect()` first."
-            )
-
-        if self.mock:
-            import tests.cameras.mock_cv2 as cv2
-        else:
-            import cv2
-
-        start_time = time.perf_counter()
-
-        frame = self.camera.wait_for_frames(timeout_ms=5000)
-
-        color_frame = frame.get_color_frame()
-
-        if not color_frame:
-            raise OSError(f"Can't capture color image from IntelRealSenseCamera({self.serial_number}).")
-
-        color_image = np.asanyarray(color_frame.get_data())
-
-        requested_color_mode = self.color_mode if temporary_color is None else temporary_color
-        if requested_color_mode not in ["rgb", "bgr"]:
-            raise ValueError(
-                f"Expected color values are 'rgb' or 'bgr', but {requested_color_mode} is provided."
-            )
-
-        # IntelRealSense uses RGB format as default (red, green, blue).
-        if requested_color_mode == "bgr":
-            color_image = cv2.cvtColor(color_image, cv2.COLOR_RGB2BGR)
-
-        h, w, _ = color_image.shape
-        if h != self.capture_height or w != self.capture_width:
-            raise OSError(
-                f"Can't capture color image with expected height and width ({self.height} x {self.width}). ({h} x {w}) returned instead."
-            )
-
-        if self.rotation is not None:
-            color_image = cv2.rotate(color_image, self.rotation)
-
-        # log the number of seconds it took to read the image
-        self.logs["delta_timestamp_s"] = time.perf_counter() - start_time
-
-        # log the utc time at which the image was received
-        self.logs["timestamp_utc"] = capture_timestamp_utc()
-
-        if self.use_depth:
-            depth_frame = frame.get_depth_frame()
-            if not depth_frame:
-                raise OSError(f"Can't capture depth image from IntelRealSenseCamera({self.serial_number}).")
-
-            depth_map = np.asanyarray(depth_frame.get_data())
-
-            h, w = depth_map.shape
-            if h != self.capture_height or w != self.capture_width:
-                raise OSError(
-                    f"Can't capture depth map with expected height and width ({self.height} x {self.width}). ({h} x {w}) returned instead."
-                )
-
-            if self.rotation is not None:
-                depth_map = cv2.rotate(depth_map, self.rotation)
-
-            return color_image, depth_map
-        else:
-            return color_image
-
-    def read_loop(self):
-        while not self.stop_event.is_set():
-            if self.use_depth:
-                self.color_image, self.depth_map = self.read()
-            else:
-                self.color_image = self.read()
-
-    def async_read(self):
-        """Access the latest color image"""
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"IntelRealSenseCamera({self.serial_number}) is not connected. Try running `camera.connect()` first."
-            )
-
-        if self.thread is None:
-            self.stop_event = threading.Event()
-            self.thread = Thread(target=self.read_loop, args=())
-            self.thread.daemon = True
-            self.thread.start()
-
-        num_tries = 0
-        while self.color_image is None:
-            # TODO(rcadene, aliberts): intelrealsense has diverged compared to opencv over here
-            num_tries += 1
-            time.sleep(1 / self.fps)
-            if num_tries > self.fps and (self.thread.ident is None or not self.thread.is_alive()):
-                raise Exception(
-                    "The thread responsible for `self.async_read()` took too much time to start. There might be an issue. Verify that `self.thread.start()` has been called."
-                )
-
-        if self.use_depth:
-            return self.color_image, self.depth_map
-        else:
-            return self.color_image
-
-    def disconnect(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"IntelRealSenseCamera({self.serial_number}) is not connected. Try running `camera.connect()` first."
-            )
-
-        if self.thread is not None and self.thread.is_alive():
-            # wait for the thread to finish
-            self.stop_event.set()
-            self.thread.join()
-            self.thread = None
-            self.stop_event = None
-
-        self.camera.stop()
-        self.camera = None
-
-        self.is_connected = False
-
-    def __del__(self):
-        if getattr(self, "is_connected", False):
-            self.disconnect()
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description="Save a few frames using `IntelRealSenseCamera` for all cameras connected to the computer, or a selected subset."
-    )
-    parser.add_argument(
-        "--serial-numbers",
-        type=int,
-        nargs="*",
-        default=None,
-        help="List of serial numbers used to instantiate the `IntelRealSenseCamera`. If not provided, find and use all available camera indices.",
-    )
-    parser.add_argument(
-        "--fps",
-        type=int,
-        default=30,
-        help="Set the number of frames recorded per seconds for all cameras. If not provided, use the default fps of each camera.",
-    )
-    parser.add_argument(
-        "--width",
-        type=int,
-        default=640,
-        help="Set the width for all cameras. If not provided, use the default width of each camera.",
-    )
-    parser.add_argument(
-        "--height",
-        type=int,
-        default=480,
-        help="Set the height for all cameras. If not provided, use the default height of each camera.",
-    )
-    parser.add_argument(
-        "--images-dir",
-        type=Path,
-        default="outputs/images_from_intelrealsense_cameras",
-        help="Set directory to save a few frames for each camera.",
-    )
-    parser.add_argument(
-        "--record-time-s",
-        type=float,
-        default=2.0,
-        help="Set the number of seconds used to record the frames. By default, 2 seconds.",
-    )
-    args = parser.parse_args()
-    save_images_from_cameras(**vars(args))

lerobot/common/robot_devices/cameras/opencv.py

@@ -1,518 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-"""
-This file contains utilities for recording frames from cameras. For more info look at `OpenCVCamera` docstring.
-"""
-
-import argparse
-import concurrent.futures
-import math
-import platform
-import shutil
-import threading
-import time
-from pathlib import Path
-from threading import Thread
-
-import numpy as np
-from PIL import Image
-
-from lerobot.common.robot_devices.cameras.configs import OpenCVCameraConfig
-from lerobot.common.robot_devices.utils import (
-    RobotDeviceAlreadyConnectedError,
-    RobotDeviceNotConnectedError,
-    busy_wait,
-)
-from lerobot.common.utils.utils import capture_timestamp_utc
-
-# The maximum opencv device index depends on your operating system. For instance,
-# if you have 3 cameras, they should be associated to index 0, 1, and 2. This is the case
-# on MacOS. However, on Ubuntu, the indices are different like 6, 16, 23.
-# When you change the USB port or reboot the computer, the operating system might
-# treat the same cameras as new devices. Thus we select a higher bound to search indices.
-MAX_OPENCV_INDEX = 60
-
-
-def find_cameras(raise_when_empty=False, max_index_search_range=MAX_OPENCV_INDEX, mock=False) -> list[dict]:
-    cameras = []
-    if platform.system() == "Linux":
-        print("Linux detected. Finding available camera indices through scanning '/dev/video*' ports")
-        possible_ports = [str(port) for port in Path("/dev").glob("video*")]
-        ports = _find_cameras(possible_ports, mock=mock)
-        for port in ports:
-            cameras.append(
-                {
-                    "port": port,
-                    "index": int(port.removeprefix("/dev/video")),
-                }
-            )
-    else:
-        print(
-            "Mac or Windows detected. Finding available camera indices through "
-            f"scanning all indices from 0 to {MAX_OPENCV_INDEX}"
-        )
-        possible_indices = range(max_index_search_range)
-        indices = _find_cameras(possible_indices, mock=mock)
-        for index in indices:
-            cameras.append(
-                {
-                    "port": None,
-                    "index": index,
-                }
-            )
-
-    return cameras
-
-
-def _find_cameras(
-    possible_camera_ids: list[int | str], raise_when_empty=False, mock=False
-) -> list[int | str]:
-    if mock:
-        import tests.cameras.mock_cv2 as cv2
-    else:
-        import cv2
-
-    camera_ids = []
-    for camera_idx in possible_camera_ids:
-        camera = cv2.VideoCapture(camera_idx)
-        is_open = camera.isOpened()
-        camera.release()
-
-        if is_open:
-            print(f"Camera found at index {camera_idx}")
-            camera_ids.append(camera_idx)
-
-    if raise_when_empty and len(camera_ids) == 0:
-        raise OSError(
-            "Not a single camera was detected. Try re-plugging, or re-installing `opencv2`, "
-            "or your camera driver, or make sure your camera is compatible with opencv2."
-        )
-
-    return camera_ids
-
-
-def is_valid_unix_path(path: str) -> bool:
-    """Note: if 'path' points to a symlink, this will return True only if the target exists"""
-    p = Path(path)
-    return p.is_absolute() and p.exists()
-
-
-def get_camera_index_from_unix_port(port: Path) -> int:
-    return int(str(port.resolve()).removeprefix("/dev/video"))
-
-
-def save_image(img_array, camera_index, frame_index, images_dir):
-    img = Image.fromarray(img_array)
-    path = images_dir / f"camera_{camera_index:02d}_frame_{frame_index:06d}.png"
-    path.parent.mkdir(parents=True, exist_ok=True)
-    img.save(str(path), quality=100)
-
-
-def save_images_from_cameras(
-    images_dir: Path,
-    camera_ids: list | None = None,
-    fps=None,
-    width=None,
-    height=None,
-    record_time_s=2,
-    mock=False,
-):
-    """
-    Initializes all the cameras and saves images to the directory. Useful to visually identify the camera
-    associated to a given camera index.
-    """
-    if camera_ids is None or len(camera_ids) == 0:
-        camera_infos = find_cameras(mock=mock)
-        camera_ids = [cam["index"] for cam in camera_infos]
-
-    print("Connecting cameras")
-    cameras = []
-    for cam_idx in camera_ids:
-        config = OpenCVCameraConfig(camera_index=cam_idx, fps=fps, width=width, height=height, mock=mock)
-        camera = OpenCVCamera(config)
-        camera.connect()
-        print(
-            f"OpenCVCamera({camera.camera_index}, fps={camera.fps}, width={camera.capture_width}, "
-            f"height={camera.capture_height}, color_mode={camera.color_mode})"
-        )
-        cameras.append(camera)
-
-    images_dir = Path(images_dir)
-    if images_dir.exists():
-        shutil.rmtree(
-            images_dir,
-        )
-    images_dir.mkdir(parents=True, exist_ok=True)
-
-    print(f"Saving images to {images_dir}")
-    frame_index = 0
-    start_time = time.perf_counter()
-    with concurrent.futures.ThreadPoolExecutor(max_workers=1) as executor:
-        while True:
-            now = time.perf_counter()
-
-            for camera in cameras:
-                # If we use async_read when fps is None, the loop will go full speed, and we will endup
-                # saving the same images from the cameras multiple times until the RAM/disk is full.
-                image = camera.read() if fps is None else camera.async_read()
-
-                executor.submit(
-                    save_image,
-                    image,
-                    camera.camera_index,
-                    frame_index,
-                    images_dir,
-                )
-
-            if fps is not None:
-                dt_s = time.perf_counter() - now
-                busy_wait(1 / fps - dt_s)
-
-            print(f"Frame: {frame_index:04d}\tLatency (ms): {(time.perf_counter() - now) * 1000:.2f}")
-
-            if time.perf_counter() - start_time > record_time_s:
-                break
-
-            frame_index += 1
-
-    print(f"Images have been saved to {images_dir}")
-
-
-class OpenCVCamera:
-    """
-    The OpenCVCamera class allows to efficiently record images from cameras. It relies on opencv2 to communicate
-    with the cameras. Most cameras are compatible. For more info, see the [Video I/O with OpenCV Overview](https://docs.opencv.org/4.x/d0/da7/videoio_overview.html).
-
-    An OpenCVCamera instance requires a camera index (e.g. `OpenCVCamera(camera_index=0)`). When you only have one camera
-    like a webcam of a laptop, the camera index is expected to be 0, but it might also be very different, and the camera index
-    might change if you reboot your computer or re-plug your camera. This behavior depends on your operation system.
-
-    To find the camera indices of your cameras, you can run our utility script that will be save a few frames for each camera:
-    ```bash
-    python lerobot/common/robot_devices/cameras/opencv.py --images-dir outputs/images_from_opencv_cameras
-    ```
-
-    When an OpenCVCamera is instantiated, if no specific config is provided, the default fps, width, height and color_mode
-    of the given camera will be used.
-
-    Example of usage:
-    ```python
-    from lerobot.common.robot_devices.cameras.configs import OpenCVCameraConfig
-
-    config = OpenCVCameraConfig(camera_index=0)
-    camera = OpenCVCamera(config)
-    camera.connect()
-    color_image = camera.read()
-    # when done using the camera, consider disconnecting
-    camera.disconnect()
-    ```
-
-    Example of changing default fps, width, height and color_mode:
-    ```python
-    config = OpenCVCameraConfig(camera_index=0, fps=30, width=1280, height=720)
-    config = OpenCVCameraConfig(camera_index=0, fps=90, width=640, height=480)
-    config = OpenCVCameraConfig(camera_index=0, fps=90, width=640, height=480, color_mode="bgr")
-    # Note: might error out open `camera.connect()` if these settings are not compatible with the camera
-    ```
-    """
-
-    def __init__(self, config: OpenCVCameraConfig):
-        self.config = config
-        self.camera_index = config.camera_index
-        self.port = None
-
-        # Linux uses ports for connecting to cameras
-        if platform.system() == "Linux":
-            if isinstance(self.camera_index, int):
-                self.port = Path(f"/dev/video{self.camera_index}")
-            elif isinstance(self.camera_index, str) and is_valid_unix_path(self.camera_index):
-                self.port = Path(self.camera_index)
-                # Retrieve the camera index from a potentially symlinked path
-                self.camera_index = get_camera_index_from_unix_port(self.port)
-            else:
-                raise ValueError(f"Please check the provided camera_index: {self.camera_index}")
-
-        # Store the raw (capture) resolution from the config.
-        self.capture_width = config.width
-        self.capture_height = config.height
-
-        # If rotated by ±90, swap width and height.
-        if config.rotation in [-90, 90]:
-            self.width = config.height
-            self.height = config.width
-        else:
-            self.width = config.width
-            self.height = config.height
-
-        self.fps = config.fps
-        self.channels = config.channels
-        self.color_mode = config.color_mode
-        self.mock = config.mock
-
-        self.camera = None
-        self.is_connected = False
-        self.thread = None
-        self.stop_event = None
-        self.color_image = None
-        self.logs = {}
-
-        if self.mock:
-            import tests.cameras.mock_cv2 as cv2
-        else:
-            import cv2
-
-        self.rotation = None
-        if config.rotation == -90:
-            self.rotation = cv2.ROTATE_90_COUNTERCLOCKWISE
-        elif config.rotation == 90:
-            self.rotation = cv2.ROTATE_90_CLOCKWISE
-        elif config.rotation == 180:
-            self.rotation = cv2.ROTATE_180
-
-    def connect(self):
-        if self.is_connected:
-            raise RobotDeviceAlreadyConnectedError(f"OpenCVCamera({self.camera_index}) is already connected.")
-
-        if self.mock:
-            import tests.cameras.mock_cv2 as cv2
-        else:
-            import cv2
-
-            # Use 1 thread to avoid blocking the main thread. Especially useful during data collection
-            # when other threads are used to save the images.
-            cv2.setNumThreads(1)
-
-        backend = (
-            cv2.CAP_V4L2
-            if platform.system() == "Linux"
-            else cv2.CAP_DSHOW
-            if platform.system() == "Windows"
-            else cv2.CAP_AVFOUNDATION
-            if platform.system() == "Darwin"
-            else cv2.CAP_ANY
-        )
-
-        camera_idx = f"/dev/video{self.camera_index}" if platform.system() == "Linux" else self.camera_index
-        # First create a temporary camera trying to access `camera_index`,
-        # and verify it is a valid camera by calling `isOpened`.
-        tmp_camera = cv2.VideoCapture(camera_idx, backend)
-        is_camera_open = tmp_camera.isOpened()
-        # Release camera to make it accessible for `find_camera_indices`
-        tmp_camera.release()
-        del tmp_camera
-
-        # If the camera doesn't work, display the camera indices corresponding to
-        # valid cameras.
-        if not is_camera_open:
-            # Verify that the provided `camera_index` is valid before printing the traceback
-            cameras_info = find_cameras()
-            available_cam_ids = [cam["index"] for cam in cameras_info]
-            if self.camera_index not in available_cam_ids:
-                raise ValueError(
-                    f"`camera_index` is expected to be one of these available cameras {available_cam_ids}, but {self.camera_index} is provided instead. "
-                    "To find the camera index you should use, run `python lerobot/common/robot_devices/cameras/opencv.py`."
-                )
-
-            raise OSError(f"Can't access OpenCVCamera({camera_idx}).")
-
-        # Secondly, create the camera that will be used downstream.
-        # Note: For some unknown reason, calling `isOpened` blocks the camera which then
-        # needs to be re-created.
-        self.camera = cv2.VideoCapture(camera_idx, backend)
-
-        if self.fps is not None:
-            self.camera.set(cv2.CAP_PROP_FPS, self.fps)
-        if self.capture_width is not None:
-            self.camera.set(cv2.CAP_PROP_FRAME_WIDTH, self.capture_width)
-        if self.capture_height is not None:
-            self.camera.set(cv2.CAP_PROP_FRAME_HEIGHT, self.capture_height)
-
-        actual_fps = self.camera.get(cv2.CAP_PROP_FPS)
-        actual_width = self.camera.get(cv2.CAP_PROP_FRAME_WIDTH)
-        actual_height = self.camera.get(cv2.CAP_PROP_FRAME_HEIGHT)
-
-        # Using `math.isclose` since actual fps can be a float (e.g. 29.9 instead of 30)
-        if self.fps is not None and not math.isclose(self.fps, actual_fps, rel_tol=1e-3):
-            # Using `OSError` since it's a broad that encompasses issues related to device communication
-            raise OSError(
-                f"Can't set {self.fps=} for OpenCVCamera({self.camera_index}). Actual value is {actual_fps}."
-            )
-        if self.capture_width is not None and not math.isclose(
-            self.capture_width, actual_width, rel_tol=1e-3
-        ):
-            raise OSError(
-                f"Can't set {self.capture_width=} for OpenCVCamera({self.camera_index}). Actual value is {actual_width}."
-            )
-        if self.capture_height is not None and not math.isclose(
-            self.capture_height, actual_height, rel_tol=1e-3
-        ):
-            raise OSError(
-                f"Can't set {self.capture_height=} for OpenCVCamera({self.camera_index}). Actual value is {actual_height}."
-            )
-
-        self.fps = round(actual_fps)
-        self.capture_width = round(actual_width)
-        self.capture_height = round(actual_height)
-        self.is_connected = True
-
-    def read(self, temporary_color_mode: str | None = None) -> np.ndarray:
-        """Read a frame from the camera returned in the format (height, width, channels)
-        (e.g. 480 x 640 x 3), contrarily to the pytorch format which is channel first.
-
-        Note: Reading a frame is done every `camera.fps` times per second, and it is blocking.
-        If you are reading data from other sensors, we advise to use `camera.async_read()` which is non blocking version of `camera.read()`.
-        """
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"OpenCVCamera({self.camera_index}) is not connected. Try running `camera.connect()` first."
-            )
-
-        start_time = time.perf_counter()
-
-        ret, color_image = self.camera.read()
-
-        if not ret:
-            raise OSError(f"Can't capture color image from camera {self.camera_index}.")
-
-        requested_color_mode = self.color_mode if temporary_color_mode is None else temporary_color_mode
-
-        if requested_color_mode not in ["rgb", "bgr"]:
-            raise ValueError(
-                f"Expected color values are 'rgb' or 'bgr', but {requested_color_mode} is provided."
-            )
-
-        # OpenCV uses BGR format as default (blue, green, red) for all operations, including displaying images.
-        # However, Deep Learning framework such as LeRobot uses RGB format as default to train neural networks,
-        # so we convert the image color from BGR to RGB.
-        if requested_color_mode == "rgb":
-            if self.mock:
-                import tests.cameras.mock_cv2 as cv2
-            else:
-                import cv2
-
-            color_image = cv2.cvtColor(color_image, cv2.COLOR_BGR2RGB)
-
-        h, w, _ = color_image.shape
-        if h != self.capture_height or w != self.capture_width:
-            raise OSError(
-                f"Can't capture color image with expected height and width ({self.height} x {self.width}). ({h} x {w}) returned instead."
-            )
-
-        if self.rotation is not None:
-            color_image = cv2.rotate(color_image, self.rotation)
-
-        # log the number of seconds it took to read the image
-        self.logs["delta_timestamp_s"] = time.perf_counter() - start_time
-
-        # log the utc time at which the image was received
-        self.logs["timestamp_utc"] = capture_timestamp_utc()
-
-        self.color_image = color_image
-
-        return color_image
-
-    def read_loop(self):
-        while not self.stop_event.is_set():
-            try:
-                self.color_image = self.read()
-            except Exception as e:
-                print(f"Error reading in thread: {e}")
-
-    def async_read(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"OpenCVCamera({self.camera_index}) is not connected. Try running `camera.connect()` first."
-            )
-
-        if self.thread is None:
-            self.stop_event = threading.Event()
-            self.thread = Thread(target=self.read_loop, args=())
-            self.thread.daemon = True
-            self.thread.start()
-
-        num_tries = 0
-        while True:
-            if self.color_image is not None:
-                return self.color_image
-
-            time.sleep(1 / self.fps)
-            num_tries += 1
-            if num_tries > self.fps * 2:
-                raise TimeoutError("Timed out waiting for async_read() to start.")
-
-    def disconnect(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"OpenCVCamera({self.camera_index}) is not connected. Try running `camera.connect()` first."
-            )
-
-        if self.thread is not None:
-            self.stop_event.set()
-            self.thread.join()  # wait for the thread to finish
-            self.thread = None
-            self.stop_event = None
-
-        self.camera.release()
-        self.camera = None
-        self.is_connected = False
-
-    def __del__(self):
-        if getattr(self, "is_connected", False):
-            self.disconnect()
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description="Save a few frames using `OpenCVCamera` for all cameras connected to the computer, or a selected subset."
-    )
-    parser.add_argument(
-        "--camera-ids",
-        type=int,
-        nargs="*",
-        default=None,
-        help="List of camera indices used to instantiate the `OpenCVCamera`. If not provided, find and use all available camera indices.",
-    )
-    parser.add_argument(
-        "--fps",
-        type=int,
-        default=None,
-        help="Set the number of frames recorded per seconds for all cameras. If not provided, use the default fps of each camera.",
-    )
-    parser.add_argument(
-        "--width",
-        type=int,
-        default=None,
-        help="Set the width for all cameras. If not provided, use the default width of each camera.",
-    )
-    parser.add_argument(
-        "--height",
-        type=int,
-        default=None,
-        help="Set the height for all cameras. If not provided, use the default height of each camera.",
-    )
-    parser.add_argument(
-        "--images-dir",
-        type=Path,
-        default="outputs/images_from_opencv_cameras",
-        help="Set directory to save a few frames for each camera.",
-    )
-    parser.add_argument(
-        "--record-time-s",
-        type=float,
-        default=4.0,
-        help="Set the number of seconds used to record the frames. By default, 2 seconds.",
-    )
-    args = parser.parse_args()
-    save_images_from_cameras(**vars(args))

lerobot/common/robot_devices/cameras/utils.py

@@ -1,67 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from typing import Protocol
-
-import numpy as np
-
-from lerobot.common.robot_devices.cameras.configs import (
-    CameraConfig,
-    IntelRealSenseCameraConfig,
-    OpenCVCameraConfig,
-)
-
-
-# Defines a camera type
-class Camera(Protocol):
-    def connect(self): ...
-    def read(self, temporary_color: str | None = None) -> np.ndarray: ...
-    def async_read(self) -> np.ndarray: ...
-    def disconnect(self): ...
-
-
-def make_cameras_from_configs(camera_configs: dict[str, CameraConfig]) -> list[Camera]:
-    cameras = {}
-
-    for key, cfg in camera_configs.items():
-        if cfg.type == "opencv":
-            from lerobot.common.robot_devices.cameras.opencv import OpenCVCamera
-
-            cameras[key] = OpenCVCamera(cfg)
-
-        elif cfg.type == "intelrealsense":
-            from lerobot.common.robot_devices.cameras.intelrealsense import IntelRealSenseCamera
-
-            cameras[key] = IntelRealSenseCamera(cfg)
-        else:
-            raise ValueError(f"The camera type '{cfg.type}' is not valid.")
-
-    return cameras
-
-
-def make_camera(camera_type, **kwargs) -> Camera:
-    if camera_type == "opencv":
-        from lerobot.common.robot_devices.cameras.opencv import OpenCVCamera
-
-        config = OpenCVCameraConfig(**kwargs)
-        return OpenCVCamera(config)
-
-    elif camera_type == "intelrealsense":
-        from lerobot.common.robot_devices.cameras.intelrealsense import IntelRealSenseCamera
-
-        config = IntelRealSenseCameraConfig(**kwargs)
-        return IntelRealSenseCamera(config)
-
-    else:
-        raise ValueError(f"The camera type '{camera_type}' is not valid.")

lerobot/common/robot_devices/control_configs.py

@@ -1,134 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from dataclasses import dataclass
-from pathlib import Path
-
-import draccus
-
-from lerobot.common.robot_devices.robots.configs import RobotConfig
-from lerobot.configs import parser
-from lerobot.configs.policies import PreTrainedConfig
-
-
-@dataclass
-class ControlConfig(draccus.ChoiceRegistry):
-    pass
-
-
-@ControlConfig.register_subclass("calibrate")
-@dataclass
-class CalibrateControlConfig(ControlConfig):
-    # List of arms to calibrate (e.g. `--arms='["left_follower","right_follower"]' left_leader`)
-    arms: list[str] | None = None
-
-
-@ControlConfig.register_subclass("teleoperate")
-@dataclass
-class TeleoperateControlConfig(ControlConfig):
-    # Limit the maximum frames per second. By default, no limit.
-    fps: int | None = None
-    teleop_time_s: float | None = None
-    # Display all cameras on screen
-    display_data: bool = False
-
-
-@ControlConfig.register_subclass("record")
-@dataclass
-class RecordControlConfig(ControlConfig):
-    # Dataset identifier. By convention it should match '{hf_username}/{dataset_name}' (e.g. `lerobot/test`).
-    repo_id: str
-    # A short but accurate description of the task performed during the recording (e.g. "Pick the Lego block and drop it in the box on the right.")
-    single_task: str
-    # Root directory where the dataset will be stored (e.g. 'dataset/path').
-    root: str | Path | None = None
-    policy: PreTrainedConfig | None = None
-    # Limit the frames per second. By default, uses the policy fps.
-    fps: int | None = None
-    # Number of seconds before starting data collection. It allows the robot devices to warmup and synchronize.
-    warmup_time_s: int | float = 10
-    # Number of seconds for data recording for each episode.
-    episode_time_s: int | float = 60
-    # Number of seconds for resetting the environment after each episode.
-    reset_time_s: int | float = 60
-    # Number of episodes to record.
-    num_episodes: int = 50
-    # Encode frames in the dataset into video
-    video: bool = True
-    # Upload dataset to Hugging Face hub.
-    push_to_hub: bool = True
-    # Upload on private repository on the Hugging Face hub.
-    private: bool = False
-    # Add tags to your dataset on the hub.
-    tags: list[str] | None = None
-    # Number of subprocesses handling the saving of frames as PNG. Set to 0 to use threads only;
-    # set to ≥1 to use subprocesses, each using threads to write images. The best number of processes
-    # and threads depends on your system. We recommend 4 threads per camera with 0 processes.
-    # If fps is unstable, adjust the thread count. If still unstable, try using 1 or more subprocesses.
-    num_image_writer_processes: int = 0
-    # Number of threads writing the frames as png images on disk, per camera.
-    # Too many threads might cause unstable teleoperation fps due to main thread being blocked.
-    # Not enough threads might cause low camera fps.
-    num_image_writer_threads_per_camera: int = 4
-    # Display all cameras on screen
-    display_data: bool = False
-    # Use vocal synthesis to read events.
-    play_sounds: bool = True
-    # Resume recording on an existing dataset.
-    resume: bool = False
-
-    def __post_init__(self):
-        # HACK: We parse again the cli args here to get the pretrained path if there was one.
-        policy_path = parser.get_path_arg("control.policy")
-        if policy_path:
-            cli_overrides = parser.get_cli_overrides("control.policy")
-            self.policy = PreTrainedConfig.from_pretrained(policy_path, cli_overrides=cli_overrides)
-            self.policy.pretrained_path = policy_path
-
-
-@ControlConfig.register_subclass("replay")
-@dataclass
-class ReplayControlConfig(ControlConfig):
-    # Dataset identifier. By convention it should match '{hf_username}/{dataset_name}' (e.g. `lerobot/test`).
-    repo_id: str
-    # Index of the episode to replay.
-    episode: int
-    # Root directory where the dataset will be stored (e.g. 'dataset/path').
-    root: str | Path | None = None
-    # Limit the frames per second. By default, uses the dataset fps.
-    fps: int | None = None
-    # Use vocal synthesis to read events.
-    play_sounds: bool = True
-
-
-@ControlConfig.register_subclass("remote_robot")
-@dataclass
-class RemoteRobotConfig(ControlConfig):
-    log_interval: int = 100
-    # Display all cameras on screen
-    display_data: bool = False
-    # Rerun configuration for remote robot (https://ref.rerun.io/docs/python/0.22.1/common/initialization_functions/#rerun.connect_tcp)
-    viewer_ip: str | None = None
-    viewer_port: str | None = None
-
-
-@dataclass
-class ControlPipelineConfig:
-    robot: RobotConfig
-    control: ControlConfig
-
-    @classmethod
-    def __get_path_fields__(cls) -> list[str]:
-        """This enables the parser to load config from the policy using `--policy.path=local/dir`"""
-        return ["control.policy"]

lerobot/common/robot_devices/motors/dynamixel.py

@@ -1,873 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import enum
-import logging
-import math
-import time
-import traceback
-from copy import deepcopy
-
-import numpy as np
-import tqdm
-
-from lerobot.common.robot_devices.motors.configs import DynamixelMotorsBusConfig
-from lerobot.common.robot_devices.utils import RobotDeviceAlreadyConnectedError, RobotDeviceNotConnectedError
-from lerobot.common.utils.utils import capture_timestamp_utc
-
-PROTOCOL_VERSION = 2.0
-BAUDRATE = 1_000_000
-TIMEOUT_MS = 1000
-
-MAX_ID_RANGE = 252
-
-# The following bounds define the lower and upper joints range (after calibration).
-# For joints in degree (i.e. revolute joints), their nominal range is [-180, 180] degrees
-# which corresponds to a half rotation on the left and half rotation on the right.
-# Some joints might require higher range, so we allow up to [-270, 270] degrees until
-# an error is raised.
-LOWER_BOUND_DEGREE = -270
-UPPER_BOUND_DEGREE = 270
-# For joints in percentage (i.e. joints that move linearly like the prismatic joint of a gripper),
-# their nominal range is [0, 100] %. For instance, for Aloha gripper, 0% is fully
-# closed, and 100% is fully open. To account for slight calibration issue, we allow up to
-# [-10, 110] until an error is raised.
-LOWER_BOUND_LINEAR = -10
-UPPER_BOUND_LINEAR = 110
-
-HALF_TURN_DEGREE = 180
-
-# https://emanual.robotis.com/docs/en/dxl/x/xl330-m077
-# https://emanual.robotis.com/docs/en/dxl/x/xl330-m288
-# https://emanual.robotis.com/docs/en/dxl/x/xl430-w250
-# https://emanual.robotis.com/docs/en/dxl/x/xm430-w350
-# https://emanual.robotis.com/docs/en/dxl/x/xm540-w270
-# https://emanual.robotis.com/docs/en/dxl/x/xc430-w150
-
-# data_name: (address, size_byte)
-X_SERIES_CONTROL_TABLE = {
-    "Model_Number": (0, 2),
-    "Model_Information": (2, 4),
-    "Firmware_Version": (6, 1),
-    "ID": (7, 1),
-    "Baud_Rate": (8, 1),
-    "Return_Delay_Time": (9, 1),
-    "Drive_Mode": (10, 1),
-    "Operating_Mode": (11, 1),
-    "Secondary_ID": (12, 1),
-    "Protocol_Type": (13, 1),
-    "Homing_Offset": (20, 4),
-    "Moving_Threshold": (24, 4),
-    "Temperature_Limit": (31, 1),
-    "Max_Voltage_Limit": (32, 2),
-    "Min_Voltage_Limit": (34, 2),
-    "PWM_Limit": (36, 2),
-    "Current_Limit": (38, 2),
-    "Acceleration_Limit": (40, 4),
-    "Velocity_Limit": (44, 4),
-    "Max_Position_Limit": (48, 4),
-    "Min_Position_Limit": (52, 4),
-    "Shutdown": (63, 1),
-    "Torque_Enable": (64, 1),
-    "LED": (65, 1),
-    "Status_Return_Level": (68, 1),
-    "Registered_Instruction": (69, 1),
-    "Hardware_Error_Status": (70, 1),
-    "Velocity_I_Gain": (76, 2),
-    "Velocity_P_Gain": (78, 2),
-    "Position_D_Gain": (80, 2),
-    "Position_I_Gain": (82, 2),
-    "Position_P_Gain": (84, 2),
-    "Feedforward_2nd_Gain": (88, 2),
-    "Feedforward_1st_Gain": (90, 2),
-    "Bus_Watchdog": (98, 1),
-    "Goal_PWM": (100, 2),
-    "Goal_Current": (102, 2),
-    "Goal_Velocity": (104, 4),
-    "Profile_Acceleration": (108, 4),
-    "Profile_Velocity": (112, 4),
-    "Goal_Position": (116, 4),
-    "Realtime_Tick": (120, 2),
-    "Moving": (122, 1),
-    "Moving_Status": (123, 1),
-    "Present_PWM": (124, 2),
-    "Present_Current": (126, 2),
-    "Present_Velocity": (128, 4),
-    "Present_Position": (132, 4),
-    "Velocity_Trajectory": (136, 4),
-    "Position_Trajectory": (140, 4),
-    "Present_Input_Voltage": (144, 2),
-    "Present_Temperature": (146, 1),
-}
-
-X_SERIES_BAUDRATE_TABLE = {
-    0: 9_600,
-    1: 57_600,
-    2: 115_200,
-    3: 1_000_000,
-    4: 2_000_000,
-    5: 3_000_000,
-    6: 4_000_000,
-}
-
-CALIBRATION_REQUIRED = ["Goal_Position", "Present_Position"]
-CONVERT_UINT32_TO_INT32_REQUIRED = ["Goal_Position", "Present_Position"]
-
-MODEL_CONTROL_TABLE = {
-    "x_series": X_SERIES_CONTROL_TABLE,
-    "xl330-m077": X_SERIES_CONTROL_TABLE,
-    "xl330-m288": X_SERIES_CONTROL_TABLE,
-    "xl430-w250": X_SERIES_CONTROL_TABLE,
-    "xm430-w350": X_SERIES_CONTROL_TABLE,
-    "xm540-w270": X_SERIES_CONTROL_TABLE,
-    "xc430-w150": X_SERIES_CONTROL_TABLE,
-}
-
-MODEL_RESOLUTION = {
-    "x_series": 4096,
-    "xl330-m077": 4096,
-    "xl330-m288": 4096,
-    "xl430-w250": 4096,
-    "xm430-w350": 4096,
-    "xm540-w270": 4096,
-    "xc430-w150": 4096,
-}
-
-MODEL_BAUDRATE_TABLE = {
-    "x_series": X_SERIES_BAUDRATE_TABLE,
-    "xl330-m077": X_SERIES_BAUDRATE_TABLE,
-    "xl330-m288": X_SERIES_BAUDRATE_TABLE,
-    "xl430-w250": X_SERIES_BAUDRATE_TABLE,
-    "xm430-w350": X_SERIES_BAUDRATE_TABLE,
-    "xm540-w270": X_SERIES_BAUDRATE_TABLE,
-    "xc430-w150": X_SERIES_BAUDRATE_TABLE,
-}
-
-NUM_READ_RETRY = 10
-NUM_WRITE_RETRY = 10
-
-
-def convert_degrees_to_steps(degrees: float | np.ndarray, models: str | list[str]) -> np.ndarray:
-    """This function converts the degree range to the step range for indicating motors rotation.
-    It assumes a motor achieves a full rotation by going from -180 degree position to +180.
-    The motor resolution (e.g. 4096) corresponds to the number of steps needed to achieve a full rotation.
-    """
-    resolutions = [MODEL_RESOLUTION[model] for model in models]
-    steps = degrees / 180 * np.array(resolutions) / 2
-    steps = steps.astype(int)
-    return steps
-
-
-def convert_to_bytes(value, bytes, mock=False):
-    if mock:
-        return value
-
-    import dynamixel_sdk as dxl
-
-    # Note: No need to convert back into unsigned int, since this byte preprocessing
-    # already handles it for us.
-    if bytes == 1:
-        data = [
-            dxl.DXL_LOBYTE(dxl.DXL_LOWORD(value)),
-        ]
-    elif bytes == 2:
-        data = [
-            dxl.DXL_LOBYTE(dxl.DXL_LOWORD(value)),
-            dxl.DXL_HIBYTE(dxl.DXL_LOWORD(value)),
-        ]
-    elif bytes == 4:
-        data = [
-            dxl.DXL_LOBYTE(dxl.DXL_LOWORD(value)),
-            dxl.DXL_HIBYTE(dxl.DXL_LOWORD(value)),
-            dxl.DXL_LOBYTE(dxl.DXL_HIWORD(value)),
-            dxl.DXL_HIBYTE(dxl.DXL_HIWORD(value)),
-        ]
-    else:
-        raise NotImplementedError(
-            f"Value of the number of bytes to be sent is expected to be in [1, 2, 4], but "
-            f"{bytes} is provided instead."
-        )
-    return data
-
-
-def get_group_sync_key(data_name, motor_names):
-    group_key = f"{data_name}_" + "_".join(motor_names)
-    return group_key
-
-
-def get_result_name(fn_name, data_name, motor_names):
-    group_key = get_group_sync_key(data_name, motor_names)
-    rslt_name = f"{fn_name}_{group_key}"
-    return rslt_name
-
-
-def get_queue_name(fn_name, data_name, motor_names):
-    group_key = get_group_sync_key(data_name, motor_names)
-    queue_name = f"{fn_name}_{group_key}"
-    return queue_name
-
-
-def get_log_name(var_name, fn_name, data_name, motor_names):
-    group_key = get_group_sync_key(data_name, motor_names)
-    log_name = f"{var_name}_{fn_name}_{group_key}"
-    return log_name
-
-
-def assert_same_address(model_ctrl_table, motor_models, data_name):
-    all_addr = []
-    all_bytes = []
-    for model in motor_models:
-        addr, bytes = model_ctrl_table[model][data_name]
-        all_addr.append(addr)
-        all_bytes.append(bytes)
-
-    if len(set(all_addr)) != 1:
-        raise NotImplementedError(
-            f"At least two motor models use a different address for `data_name`='{data_name}' ({list(zip(motor_models, all_addr, strict=False))}). Contact a LeRobot maintainer."
-        )
-
-    if len(set(all_bytes)) != 1:
-        raise NotImplementedError(
-            f"At least two motor models use a different bytes representation for `data_name`='{data_name}' ({list(zip(motor_models, all_bytes, strict=False))}). Contact a LeRobot maintainer."
-        )
-
-
-class TorqueMode(enum.Enum):
-    ENABLED = 1
-    DISABLED = 0
-
-
-class DriveMode(enum.Enum):
-    NON_INVERTED = 0
-    INVERTED = 1
-
-
-class CalibrationMode(enum.Enum):
-    # Joints with rotational motions are expressed in degrees in nominal range of [-180, 180]
-    DEGREE = 0
-    # Joints with linear motions (like gripper of Aloha) are expressed in nominal range of [0, 100]
-    LINEAR = 1
-
-
-class JointOutOfRangeError(Exception):
-    def __init__(self, message="Joint is out of range"):
-        self.message = message
-        super().__init__(self.message)
-
-
-class DynamixelMotorsBus:
-    """
-    The DynamixelMotorsBus class allows to efficiently read and write to the attached motors. It relies on
-    the python dynamixel sdk to communicate with the motors. For more info, see the [Dynamixel SDK Documentation](https://emanual.robotis.com/docs/en/software/dynamixel/dynamixel_sdk/sample_code/python_read_write_protocol_2_0/#python-read-write-protocol-20).
-
-    A DynamixelMotorsBus instance requires a port (e.g. `DynamixelMotorsBus(port="/dev/tty.usbmodem575E0031751"`)).
-    To find the port, you can run our utility script:
-    ```bash
-    python lerobot/scripts/find_motors_bus_port.py
-    >>> Finding all available ports for the MotorBus.
-    >>> ['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-    >>> Remove the usb cable from your DynamixelMotorsBus and press Enter when done.
-    >>> The port of this DynamixelMotorsBus is /dev/tty.usbmodem575E0031751.
-    >>> Reconnect the usb cable.
-    ```
-
-    Example of usage for 1 motor connected to the bus:
-    ```python
-    motor_name = "gripper"
-    motor_index = 6
-    motor_model = "xl330-m288"
-
-    config = DynamixelMotorsBusConfig(
-        port="/dev/tty.usbmodem575E0031751",
-        motors={motor_name: (motor_index, motor_model)},
-    )
-    motors_bus = DynamixelMotorsBus(config)
-    motors_bus.connect()
-
-    position = motors_bus.read("Present_Position")
-
-    # move from a few motor steps as an example
-    few_steps = 30
-    motors_bus.write("Goal_Position", position + few_steps)
-
-    # when done, consider disconnecting
-    motors_bus.disconnect()
-    ```
-    """
-
-    def __init__(
-        self,
-        config: DynamixelMotorsBusConfig,
-    ):
-        self.port = config.port
-        self.motors = config.motors
-        self.mock = config.mock
-
-        self.model_ctrl_table = deepcopy(MODEL_CONTROL_TABLE)
-        self.model_resolution = deepcopy(MODEL_RESOLUTION)
-
-        self.port_handler = None
-        self.packet_handler = None
-        self.calibration = None
-        self.is_connected = False
-        self.group_readers = {}
-        self.group_writers = {}
-        self.logs = {}
-
-    def connect(self):
-        if self.is_connected:
-            raise RobotDeviceAlreadyConnectedError(
-                f"DynamixelMotorsBus({self.port}) is already connected. Do not call `motors_bus.connect()` twice."
-            )
-
-        if self.mock:
-            import tests.motors.mock_dynamixel_sdk as dxl
-        else:
-            import dynamixel_sdk as dxl
-
-        self.port_handler = dxl.PortHandler(self.port)
-        self.packet_handler = dxl.PacketHandler(PROTOCOL_VERSION)
-
-        try:
-            if not self.port_handler.openPort():
-                raise OSError(f"Failed to open port '{self.port}'.")
-        except Exception:
-            traceback.print_exc()
-            print(
-                "\nTry running `python lerobot/scripts/find_motors_bus_port.py` to make sure you are using the correct port.\n"
-            )
-            raise
-
-        # Allow to read and write
-        self.is_connected = True
-
-        self.port_handler.setPacketTimeoutMillis(TIMEOUT_MS)
-
-    def reconnect(self):
-        if self.mock:
-            import tests.motors.mock_dynamixel_sdk as dxl
-        else:
-            import dynamixel_sdk as dxl
-
-        self.port_handler = dxl.PortHandler(self.port)
-        self.packet_handler = dxl.PacketHandler(PROTOCOL_VERSION)
-
-        if not self.port_handler.openPort():
-            raise OSError(f"Failed to open port '{self.port}'.")
-
-        self.is_connected = True
-
-    def are_motors_configured(self):
-        # Only check the motor indices and not baudrate, since if the motor baudrates are incorrect,
-        # a ConnectionError will be raised anyway.
-        try:
-            return (self.motor_indices == self.read("ID")).all()
-        except ConnectionError as e:
-            print(e)
-            return False
-
-    def find_motor_indices(self, possible_ids=None, num_retry=2):
-        if possible_ids is None:
-            possible_ids = range(MAX_ID_RANGE)
-
-        indices = []
-        for idx in tqdm.tqdm(possible_ids):
-            try:
-                present_idx = self.read_with_motor_ids(self.motor_models, [idx], "ID", num_retry=num_retry)[0]
-            except ConnectionError:
-                continue
-
-            if idx != present_idx:
-                # sanity check
-                raise OSError(
-                    "Motor index used to communicate through the bus is not the same as the one present in the motor memory. The motor memory might be damaged."
-                )
-            indices.append(idx)
-
-        return indices
-
-    def set_bus_baudrate(self, baudrate):
-        present_bus_baudrate = self.port_handler.getBaudRate()
-        if present_bus_baudrate != baudrate:
-            print(f"Setting bus baud rate to {baudrate}. Previously {present_bus_baudrate}.")
-            self.port_handler.setBaudRate(baudrate)
-
-            if self.port_handler.getBaudRate() != baudrate:
-                raise OSError("Failed to write bus baud rate.")
-
-    @property
-    def motor_names(self) -> list[str]:
-        return list(self.motors.keys())
-
-    @property
-    def motor_models(self) -> list[str]:
-        return [model for _, model in self.motors.values()]
-
-    @property
-    def motor_indices(self) -> list[int]:
-        return [idx for idx, _ in self.motors.values()]
-
-    def set_calibration(self, calibration: dict[str, list]):
-        self.calibration = calibration
-
-    def apply_calibration_autocorrect(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """This function applies the calibration, automatically detects out of range errors for motors values and attempts to correct.
-
-        For more info, see docstring of `apply_calibration` and `autocorrect_calibration`.
-        """
-        try:
-            values = self.apply_calibration(values, motor_names)
-        except JointOutOfRangeError as e:
-            print(e)
-            self.autocorrect_calibration(values, motor_names)
-            values = self.apply_calibration(values, motor_names)
-        return values
-
-    def apply_calibration(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """Convert from unsigned int32 joint position range [0, 2**32[ to the universal float32 nominal degree range ]-180.0, 180.0[ with
-        a "zero position" at 0 degree.
-
-        Note: We say "nominal degree range" since the motors can take values outside this range. For instance, 190 degrees, if the motor
-        rotate more than a half a turn from the zero position. However, most motors can't rotate more than 180 degrees and will stay in this range.
-
-        Joints values are original in [0, 2**32[ (unsigned int32). Each motor are expected to complete a full rotation
-        when given a goal position that is + or - their resolution. For instance, dynamixel xl330-m077 have a resolution of 4096, and
-        at any position in their original range, let's say the position 56734, they complete a full rotation clockwise by moving to 60830,
-        or anticlockwise by moving to 52638. The position in the original range is arbitrary and might change a lot between each motor.
-        To harmonize between motors of the same model, different robots, or even models of different brands, we propose to work
-        in the centered nominal degree range ]-180, 180[.
-        """
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        # Convert from unsigned int32 original range [0, 2**32] to signed float32 range
-        values = values.astype(np.float32)
-
-        for i, name in enumerate(motor_names):
-            calib_idx = self.calibration["motor_names"].index(name)
-            calib_mode = self.calibration["calib_mode"][calib_idx]
-
-            if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                drive_mode = self.calibration["drive_mode"][calib_idx]
-                homing_offset = self.calibration["homing_offset"][calib_idx]
-                _, model = self.motors[name]
-                resolution = self.model_resolution[model]
-
-                # Update direction of rotation of the motor to match between leader and follower.
-                # In fact, the motor of the leader for a given joint can be assembled in an
-                # opposite direction in term of rotation than the motor of the follower on the same joint.
-                if drive_mode:
-                    values[i] *= -1
-
-                # Convert from range [-2**31, 2**31] to
-                # nominal range [-resolution//2, resolution//2] (e.g. [-2048, 2048])
-                values[i] += homing_offset
-
-                # Convert from range [-resolution//2, resolution//2] to
-                # universal float32 centered degree range [-180, 180]
-                # (e.g. 2048 / (4096 // 2) * 180 = 180)
-                values[i] = values[i] / (resolution // 2) * HALF_TURN_DEGREE
-
-                if (values[i] < LOWER_BOUND_DEGREE) or (values[i] > UPPER_BOUND_DEGREE):
-                    raise JointOutOfRangeError(
-                        f"Wrong motor position range detected for {name}. "
-                        f"Expected to be in nominal range of [-{HALF_TURN_DEGREE}, {HALF_TURN_DEGREE}] degrees (a full rotation), "
-                        f"with a maximum range of [{LOWER_BOUND_DEGREE}, {UPPER_BOUND_DEGREE}] degrees to account for joints that can rotate a bit more, "
-                        f"but present value is {values[i]} degree. "
-                        "This might be due to a cable connection issue creating an artificial 360 degrees jump in motor values. "
-                        "You need to recalibrate by running: `python lerobot/scripts/control_robot.py calibrate`"
-                    )
-
-            elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                start_pos = self.calibration["start_pos"][calib_idx]
-                end_pos = self.calibration["end_pos"][calib_idx]
-
-                # Rescale the present position to a nominal range [0, 100] %,
-                # useful for joints with linear motions like Aloha gripper
-                values[i] = (values[i] - start_pos) / (end_pos - start_pos) * 100
-
-                if (values[i] < LOWER_BOUND_LINEAR) or (values[i] > UPPER_BOUND_LINEAR):
-                    raise JointOutOfRangeError(
-                        f"Wrong motor position range detected for {name}. "
-                        f"Expected to be in nominal range of [0, 100] % (a full linear translation), "
-                        f"with a maximum range of [{LOWER_BOUND_LINEAR}, {UPPER_BOUND_LINEAR}] % to account for some imprecision during calibration, "
-                        f"but present value is {values[i]} %. "
-                        "This might be due to a cable connection issue creating an artificial jump in motor values. "
-                        "You need to recalibrate by running: `python lerobot/scripts/control_robot.py calibrate`"
-                    )
-
-        return values
-
-    def autocorrect_calibration(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """This function automatically detects issues with values of motors after calibration, and correct for these issues.
-
-        Some motors might have values outside of expected maximum bounds after calibration.
-        For instance, for a joint in degree, its value can be outside [-270, 270] degrees, which is totally unexpected given
-        a nominal range of [-180, 180] degrees, which represents half a turn to the left or right starting from zero position.
-
-        Known issues:
-        #1: Motor value randomly shifts of a full turn, caused by hardware/connection errors.
-        #2: Motor internal homing offset is shifted by a full turn, caused by using default calibration (e.g Aloha).
-        #3: motor internal homing offset is shifted by less or more than a full turn, caused by using default calibration
-            or by human error during manual calibration.
-
-        Issues #1 and #2 can be solved by shifting the calibration homing offset by a full turn.
-        Issue #3 will be visually detected by user and potentially captured by the safety feature `max_relative_target`,
-        that will slow down the motor, raise an error asking to recalibrate. Manual recalibrating will solve the issue.
-
-        Note: A full turn corresponds to 360 degrees but also to 4096 steps for a motor resolution of 4096.
-        """
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        # Convert from unsigned int32 original range [0, 2**32] to signed float32 range
-        values = values.astype(np.float32)
-
-        for i, name in enumerate(motor_names):
-            calib_idx = self.calibration["motor_names"].index(name)
-            calib_mode = self.calibration["calib_mode"][calib_idx]
-
-            if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                drive_mode = self.calibration["drive_mode"][calib_idx]
-                homing_offset = self.calibration["homing_offset"][calib_idx]
-                _, model = self.motors[name]
-                resolution = self.model_resolution[model]
-
-                # Update direction of rotation of the motor to match between leader and follower.
-                # In fact, the motor of the leader for a given joint can be assembled in an
-                # opposite direction in term of rotation than the motor of the follower on the same joint.
-                if drive_mode:
-                    values[i] *= -1
-
-                # Convert from initial range to range [-180, 180] degrees
-                calib_val = (values[i] + homing_offset) / (resolution // 2) * HALF_TURN_DEGREE
-                in_range = (calib_val > LOWER_BOUND_DEGREE) and (calib_val < UPPER_BOUND_DEGREE)
-
-                # Solve this inequality to find the factor to shift the range into [-180, 180] degrees
-                # values[i] = (values[i] + homing_offset + resolution * factor) / (resolution // 2) * HALF_TURN_DEGREE
-                # - HALF_TURN_DEGREE <= (values[i] + homing_offset + resolution * factor) / (resolution // 2) * HALF_TURN_DEGREE <= HALF_TURN_DEGREE
-                # (- (resolution // 2) - values[i] - homing_offset) / resolution <= factor <= ((resolution // 2) - values[i] - homing_offset) / resolution
-                low_factor = (-(resolution // 2) - values[i] - homing_offset) / resolution
-                upp_factor = ((resolution // 2) - values[i] - homing_offset) / resolution
-
-            elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                start_pos = self.calibration["start_pos"][calib_idx]
-                end_pos = self.calibration["end_pos"][calib_idx]
-
-                # Convert from initial range to range [0, 100] in %
-                calib_val = (values[i] - start_pos) / (end_pos - start_pos) * 100
-                in_range = (calib_val > LOWER_BOUND_LINEAR) and (calib_val < UPPER_BOUND_LINEAR)
-
-                # Solve this inequality to find the factor to shift the range into [0, 100] %
-                # values[i] = (values[i] - start_pos + resolution * factor) / (end_pos + resolution * factor - start_pos - resolution * factor) * 100
-                # values[i] = (values[i] - start_pos + resolution * factor) / (end_pos - start_pos) * 100
-                # 0 <= (values[i] - start_pos + resolution * factor) / (end_pos - start_pos) * 100 <= 100
-                # (start_pos - values[i]) / resolution <= factor <= (end_pos - values[i]) / resolution
-                low_factor = (start_pos - values[i]) / resolution
-                upp_factor = (end_pos - values[i]) / resolution
-
-            if not in_range:
-                # Get first integer between the two bounds
-                if low_factor < upp_factor:
-                    factor = math.ceil(low_factor)
-
-                    if factor > upp_factor:
-                        raise ValueError(f"No integer found between bounds [{low_factor=}, {upp_factor=}]")
-                else:
-                    factor = math.ceil(upp_factor)
-
-                    if factor > low_factor:
-                        raise ValueError(f"No integer found between bounds [{low_factor=}, {upp_factor=}]")
-
-                if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                    out_of_range_str = f"{LOWER_BOUND_DEGREE} < {calib_val} < {UPPER_BOUND_DEGREE} degrees"
-                    in_range_str = f"{LOWER_BOUND_DEGREE} < {calib_val} < {UPPER_BOUND_DEGREE} degrees"
-                elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                    out_of_range_str = f"{LOWER_BOUND_LINEAR} < {calib_val} < {UPPER_BOUND_LINEAR} %"
-                    in_range_str = f"{LOWER_BOUND_LINEAR} < {calib_val} < {UPPER_BOUND_LINEAR} %"
-
-                logging.warning(
-                    f"Auto-correct calibration of motor '{name}' by shifting value by {abs(factor)} full turns, "
-                    f"from '{out_of_range_str}' to '{in_range_str}'."
-                )
-
-                # A full turn corresponds to 360 degrees but also to 4096 steps for a motor resolution of 4096.
-                self.calibration["homing_offset"][calib_idx] += resolution * factor
-
-    def revert_calibration(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """Inverse of `apply_calibration`."""
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        for i, name in enumerate(motor_names):
-            calib_idx = self.calibration["motor_names"].index(name)
-            calib_mode = self.calibration["calib_mode"][calib_idx]
-
-            if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                drive_mode = self.calibration["drive_mode"][calib_idx]
-                homing_offset = self.calibration["homing_offset"][calib_idx]
-                _, model = self.motors[name]
-                resolution = self.model_resolution[model]
-
-                # Convert from nominal 0-centered degree range [-180, 180] to
-                # 0-centered resolution range (e.g. [-2048, 2048] for resolution=4096)
-                values[i] = values[i] / HALF_TURN_DEGREE * (resolution // 2)
-
-                # Subtract the homing offsets to come back to actual motor range of values
-                # which can be arbitrary.
-                values[i] -= homing_offset
-
-                # Remove drive mode, which is the rotation direction of the motor, to come back to
-                # actual motor rotation direction which can be arbitrary.
-                if drive_mode:
-                    values[i] *= -1
-
-            elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                start_pos = self.calibration["start_pos"][calib_idx]
-                end_pos = self.calibration["end_pos"][calib_idx]
-
-                # Convert from nominal lnear range of [0, 100] % to
-                # actual motor range of values which can be arbitrary.
-                values[i] = values[i] / 100 * (end_pos - start_pos) + start_pos
-
-        values = np.round(values).astype(np.int32)
-        return values
-
-    def read_with_motor_ids(self, motor_models, motor_ids, data_name, num_retry=NUM_READ_RETRY):
-        if self.mock:
-            import tests.motors.mock_dynamixel_sdk as dxl
-        else:
-            import dynamixel_sdk as dxl
-
-        return_list = True
-        if not isinstance(motor_ids, list):
-            return_list = False
-            motor_ids = [motor_ids]
-
-        assert_same_address(self.model_ctrl_table, self.motor_models, data_name)
-        addr, bytes = self.model_ctrl_table[motor_models[0]][data_name]
-        group = dxl.GroupSyncRead(self.port_handler, self.packet_handler, addr, bytes)
-        for idx in motor_ids:
-            group.addParam(idx)
-
-        for _ in range(num_retry):
-            comm = group.txRxPacket()
-            if comm == dxl.COMM_SUCCESS:
-                break
-
-        if comm != dxl.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Read failed due to communication error on port {self.port_handler.port_name} for indices {motor_ids}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-        values = []
-        for idx in motor_ids:
-            value = group.getData(idx, addr, bytes)
-            values.append(value)
-
-        if return_list:
-            return values
-        else:
-            return values[0]
-
-    def read(self, data_name, motor_names: str | list[str] | None = None):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"DynamixelMotorsBus({self.port}) is not connected. You need to run `motors_bus.connect()`."
-            )
-
-        start_time = time.perf_counter()
-
-        if self.mock:
-            import tests.motors.mock_dynamixel_sdk as dxl
-        else:
-            import dynamixel_sdk as dxl
-
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        if isinstance(motor_names, str):
-            motor_names = [motor_names]
-
-        motor_ids = []
-        models = []
-        for name in motor_names:
-            motor_idx, model = self.motors[name]
-            motor_ids.append(motor_idx)
-            models.append(model)
-
-        assert_same_address(self.model_ctrl_table, models, data_name)
-        addr, bytes = self.model_ctrl_table[model][data_name]
-        group_key = get_group_sync_key(data_name, motor_names)
-
-        if data_name not in self.group_readers:
-            # create new group reader
-            self.group_readers[group_key] = dxl.GroupSyncRead(
-                self.port_handler, self.packet_handler, addr, bytes
-            )
-            for idx in motor_ids:
-                self.group_readers[group_key].addParam(idx)
-
-        for _ in range(NUM_READ_RETRY):
-            comm = self.group_readers[group_key].txRxPacket()
-            if comm == dxl.COMM_SUCCESS:
-                break
-
-        if comm != dxl.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Read failed due to communication error on port {self.port} for group_key {group_key}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-        values = []
-        for idx in motor_ids:
-            value = self.group_readers[group_key].getData(idx, addr, bytes)
-            values.append(value)
-
-        values = np.array(values)
-
-        # Convert to signed int to use range [-2048, 2048] for our motor positions.
-        if data_name in CONVERT_UINT32_TO_INT32_REQUIRED:
-            values = values.astype(np.int32)
-
-        if data_name in CALIBRATION_REQUIRED and self.calibration is not None:
-            values = self.apply_calibration_autocorrect(values, motor_names)
-
-        # log the number of seconds it took to read the data from the motors
-        delta_ts_name = get_log_name("delta_timestamp_s", "read", data_name, motor_names)
-        self.logs[delta_ts_name] = time.perf_counter() - start_time
-
-        # log the utc time at which the data was received
-        ts_utc_name = get_log_name("timestamp_utc", "read", data_name, motor_names)
-        self.logs[ts_utc_name] = capture_timestamp_utc()
-
-        return values
-
-    def write_with_motor_ids(self, motor_models, motor_ids, data_name, values, num_retry=NUM_WRITE_RETRY):
-        if self.mock:
-            import tests.motors.mock_dynamixel_sdk as dxl
-        else:
-            import dynamixel_sdk as dxl
-
-        if not isinstance(motor_ids, list):
-            motor_ids = [motor_ids]
-        if not isinstance(values, list):
-            values = [values]
-
-        assert_same_address(self.model_ctrl_table, motor_models, data_name)
-        addr, bytes = self.model_ctrl_table[motor_models[0]][data_name]
-        group = dxl.GroupSyncWrite(self.port_handler, self.packet_handler, addr, bytes)
-        for idx, value in zip(motor_ids, values, strict=True):
-            data = convert_to_bytes(value, bytes, self.mock)
-            group.addParam(idx, data)
-
-        for _ in range(num_retry):
-            comm = group.txPacket()
-            if comm == dxl.COMM_SUCCESS:
-                break
-
-        if comm != dxl.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Write failed due to communication error on port {self.port_handler.port_name} for indices {motor_ids}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-    def write(self, data_name, values: int | float | np.ndarray, motor_names: str | list[str] | None = None):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"DynamixelMotorsBus({self.port}) is not connected. You need to run `motors_bus.connect()`."
-            )
-
-        start_time = time.perf_counter()
-
-        if self.mock:
-            import tests.motors.mock_dynamixel_sdk as dxl
-        else:
-            import dynamixel_sdk as dxl
-
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        if isinstance(motor_names, str):
-            motor_names = [motor_names]
-
-        if isinstance(values, (int, float, np.integer)):
-            values = [int(values)] * len(motor_names)
-
-        values = np.array(values)
-
-        motor_ids = []
-        models = []
-        for name in motor_names:
-            motor_idx, model = self.motors[name]
-            motor_ids.append(motor_idx)
-            models.append(model)
-
-        if data_name in CALIBRATION_REQUIRED and self.calibration is not None:
-            values = self.revert_calibration(values, motor_names)
-
-        values = values.tolist()
-
-        assert_same_address(self.model_ctrl_table, models, data_name)
-        addr, bytes = self.model_ctrl_table[model][data_name]
-        group_key = get_group_sync_key(data_name, motor_names)
-
-        init_group = data_name not in self.group_readers
-        if init_group:
-            self.group_writers[group_key] = dxl.GroupSyncWrite(
-                self.port_handler, self.packet_handler, addr, bytes
-            )
-
-        for idx, value in zip(motor_ids, values, strict=True):
-            data = convert_to_bytes(value, bytes, self.mock)
-            if init_group:
-                self.group_writers[group_key].addParam(idx, data)
-            else:
-                self.group_writers[group_key].changeParam(idx, data)
-
-        comm = self.group_writers[group_key].txPacket()
-        if comm != dxl.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Write failed due to communication error on port {self.port} for group_key {group_key}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-        # log the number of seconds it took to write the data to the motors
-        delta_ts_name = get_log_name("delta_timestamp_s", "write", data_name, motor_names)
-        self.logs[delta_ts_name] = time.perf_counter() - start_time
-
-        # TODO(rcadene): should we log the time before sending the write command?
-        # log the utc time when the write has been completed
-        ts_utc_name = get_log_name("timestamp_utc", "write", data_name, motor_names)
-        self.logs[ts_utc_name] = capture_timestamp_utc()
-
-    def disconnect(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"DynamixelMotorsBus({self.port}) is not connected. Try running `motors_bus.connect()` first."
-            )
-
-        if self.port_handler is not None:
-            self.port_handler.closePort()
-            self.port_handler = None
-
-        self.packet_handler = None
-        self.group_readers = {}
-        self.group_writers = {}
-        self.is_connected = False
-
-    def __del__(self):
-        if getattr(self, "is_connected", False):
-            self.disconnect()

lerobot/common/robot_devices/motors/feetech.py

@@ -1,898 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import enum
-import logging
-import math
-import time
-import traceback
-from copy import deepcopy
-
-import numpy as np
-import tqdm
-
-from lerobot.common.robot_devices.motors.configs import FeetechMotorsBusConfig
-from lerobot.common.robot_devices.utils import RobotDeviceAlreadyConnectedError, RobotDeviceNotConnectedError
-from lerobot.common.utils.utils import capture_timestamp_utc
-
-PROTOCOL_VERSION = 0
-BAUDRATE = 1_000_000
-TIMEOUT_MS = 1000
-
-MAX_ID_RANGE = 252
-
-# The following bounds define the lower and upper joints range (after calibration).
-# For joints in degree (i.e. revolute joints), their nominal range is [-180, 180] degrees
-# which corresponds to a half rotation on the left and half rotation on the right.
-# Some joints might require higher range, so we allow up to [-270, 270] degrees until
-# an error is raised.
-LOWER_BOUND_DEGREE = -270
-UPPER_BOUND_DEGREE = 270
-# For joints in percentage (i.e. joints that move linearly like the prismatic joint of a gripper),
-# their nominal range is [0, 100] %. For instance, for Aloha gripper, 0% is fully
-# closed, and 100% is fully open. To account for slight calibration issue, we allow up to
-# [-10, 110] until an error is raised.
-LOWER_BOUND_LINEAR = -10
-UPPER_BOUND_LINEAR = 110
-
-HALF_TURN_DEGREE = 180
-
-
-# See this link for STS3215 Memory Table:
-# https://docs.google.com/spreadsheets/d/1GVs7W1VS1PqdhA1nW-abeyAHhTUxKUdR/edit?usp=sharing&ouid=116566590112741600240&rtpof=true&sd=true
-# data_name: (address, size_byte)
-SCS_SERIES_CONTROL_TABLE = {
-    "Model": (3, 2),
-    "ID": (5, 1),
-    "Baud_Rate": (6, 1),
-    "Return_Delay": (7, 1),
-    "Response_Status_Level": (8, 1),
-    "Min_Angle_Limit": (9, 2),
-    "Max_Angle_Limit": (11, 2),
-    "Max_Temperature_Limit": (13, 1),
-    "Max_Voltage_Limit": (14, 1),
-    "Min_Voltage_Limit": (15, 1),
-    "Max_Torque_Limit": (16, 2),
-    "Phase": (18, 1),
-    "Unloading_Condition": (19, 1),
-    "LED_Alarm_Condition": (20, 1),
-    "P_Coefficient": (21, 1),
-    "D_Coefficient": (22, 1),
-    "I_Coefficient": (23, 1),
-    "Minimum_Startup_Force": (24, 2),
-    "CW_Dead_Zone": (26, 1),
-    "CCW_Dead_Zone": (27, 1),
-    "Protection_Current": (28, 2),
-    "Angular_Resolution": (30, 1),
-    "Offset": (31, 2),
-    "Mode": (33, 1),
-    "Protective_Torque": (34, 1),
-    "Protection_Time": (35, 1),
-    "Overload_Torque": (36, 1),
-    "Speed_closed_loop_P_proportional_coefficient": (37, 1),
-    "Over_Current_Protection_Time": (38, 1),
-    "Velocity_closed_loop_I_integral_coefficient": (39, 1),
-    "Torque_Enable": (40, 1),
-    "Acceleration": (41, 1),
-    "Goal_Position": (42, 2),
-    "Goal_Time": (44, 2),
-    "Goal_Speed": (46, 2),
-    "Torque_Limit": (48, 2),
-    "Lock": (55, 1),
-    "Present_Position": (56, 2),
-    "Present_Speed": (58, 2),
-    "Present_Load": (60, 2),
-    "Present_Voltage": (62, 1),
-    "Present_Temperature": (63, 1),
-    "Status": (65, 1),
-    "Moving": (66, 1),
-    "Present_Current": (69, 2),
-    # Not in the Memory Table
-    "Maximum_Acceleration": (85, 2),
-}
-
-SCS_SERIES_BAUDRATE_TABLE = {
-    0: 1_000_000,
-    1: 500_000,
-    2: 250_000,
-    3: 128_000,
-    4: 115_200,
-    5: 57_600,
-    6: 38_400,
-    7: 19_200,
-}
-
-CALIBRATION_REQUIRED = ["Goal_Position", "Present_Position"]
-CONVERT_UINT32_TO_INT32_REQUIRED = ["Goal_Position", "Present_Position"]
-
-
-MODEL_CONTROL_TABLE = {
-    "scs_series": SCS_SERIES_CONTROL_TABLE,
-    "sts3215": SCS_SERIES_CONTROL_TABLE,
-}
-
-MODEL_RESOLUTION = {
-    "scs_series": 4096,
-    "sts3215": 4096,
-}
-
-MODEL_BAUDRATE_TABLE = {
-    "scs_series": SCS_SERIES_BAUDRATE_TABLE,
-    "sts3215": SCS_SERIES_BAUDRATE_TABLE,
-}
-
-# High number of retries is needed for feetech compared to dynamixel motors.
-NUM_READ_RETRY = 20
-NUM_WRITE_RETRY = 20
-
-
-def convert_degrees_to_steps(degrees: float | np.ndarray, models: str | list[str]) -> np.ndarray:
-    """This function converts the degree range to the step range for indicating motors rotation.
-    It assumes a motor achieves a full rotation by going from -180 degree position to +180.
-    The motor resolution (e.g. 4096) corresponds to the number of steps needed to achieve a full rotation.
-    """
-    resolutions = [MODEL_RESOLUTION[model] for model in models]
-    steps = degrees / 180 * np.array(resolutions) / 2
-    steps = steps.astype(int)
-    return steps
-
-
-def convert_to_bytes(value, bytes, mock=False):
-    if mock:
-        return value
-
-    import scservo_sdk as scs
-
-    # Note: No need to convert back into unsigned int, since this byte preprocessing
-    # already handles it for us.
-    if bytes == 1:
-        data = [
-            scs.SCS_LOBYTE(scs.SCS_LOWORD(value)),
-        ]
-    elif bytes == 2:
-        data = [
-            scs.SCS_LOBYTE(scs.SCS_LOWORD(value)),
-            scs.SCS_HIBYTE(scs.SCS_LOWORD(value)),
-        ]
-    elif bytes == 4:
-        data = [
-            scs.SCS_LOBYTE(scs.SCS_LOWORD(value)),
-            scs.SCS_HIBYTE(scs.SCS_LOWORD(value)),
-            scs.SCS_LOBYTE(scs.SCS_HIWORD(value)),
-            scs.SCS_HIBYTE(scs.SCS_HIWORD(value)),
-        ]
-    else:
-        raise NotImplementedError(
-            f"Value of the number of bytes to be sent is expected to be in [1, 2, 4], but "
-            f"{bytes} is provided instead."
-        )
-    return data
-
-
-def get_group_sync_key(data_name, motor_names):
-    group_key = f"{data_name}_" + "_".join(motor_names)
-    return group_key
-
-
-def get_result_name(fn_name, data_name, motor_names):
-    group_key = get_group_sync_key(data_name, motor_names)
-    rslt_name = f"{fn_name}_{group_key}"
-    return rslt_name
-
-
-def get_queue_name(fn_name, data_name, motor_names):
-    group_key = get_group_sync_key(data_name, motor_names)
-    queue_name = f"{fn_name}_{group_key}"
-    return queue_name
-
-
-def get_log_name(var_name, fn_name, data_name, motor_names):
-    group_key = get_group_sync_key(data_name, motor_names)
-    log_name = f"{var_name}_{fn_name}_{group_key}"
-    return log_name
-
-
-def assert_same_address(model_ctrl_table, motor_models, data_name):
-    all_addr = []
-    all_bytes = []
-    for model in motor_models:
-        addr, bytes = model_ctrl_table[model][data_name]
-        all_addr.append(addr)
-        all_bytes.append(bytes)
-
-    if len(set(all_addr)) != 1:
-        raise NotImplementedError(
-            f"At least two motor models use a different address for `data_name`='{data_name}' ({list(zip(motor_models, all_addr, strict=False))}). Contact a LeRobot maintainer."
-        )
-
-    if len(set(all_bytes)) != 1:
-        raise NotImplementedError(
-            f"At least two motor models use a different bytes representation for `data_name`='{data_name}' ({list(zip(motor_models, all_bytes, strict=False))}). Contact a LeRobot maintainer."
-        )
-
-
-class TorqueMode(enum.Enum):
-    ENABLED = 1
-    DISABLED = 0
-
-
-class DriveMode(enum.Enum):
-    NON_INVERTED = 0
-    INVERTED = 1
-
-
-class CalibrationMode(enum.Enum):
-    # Joints with rotational motions are expressed in degrees in nominal range of [-180, 180]
-    DEGREE = 0
-    # Joints with linear motions (like gripper of Aloha) are expressed in nominal range of [0, 100]
-    LINEAR = 1
-
-
-class JointOutOfRangeError(Exception):
-    def __init__(self, message="Joint is out of range"):
-        self.message = message
-        super().__init__(self.message)
-
-
-class FeetechMotorsBus:
-    """
-    The FeetechMotorsBus class allows to efficiently read and write to the attached motors. It relies on
-    the python feetech sdk to communicate with the motors. For more info, see the [feetech SDK Documentation](https://emanual.robotis.com/docs/en/software/feetech/feetech_sdk/sample_code/python_read_write_protocol_2_0/#python-read-write-protocol-20).
-
-    A FeetechMotorsBus instance requires a port (e.g. `FeetechMotorsBus(port="/dev/tty.usbmodem575E0031751"`)).
-    To find the port, you can run our utility script:
-    ```bash
-    python lerobot/scripts/find_motors_bus_port.py
-    >>> Finding all available ports for the MotorsBus.
-    >>> ['/dev/tty.usbmodem575E0032081', '/dev/tty.usbmodem575E0031751']
-    >>> Remove the usb cable from your FeetechMotorsBus and press Enter when done.
-    >>> The port of this FeetechMotorsBus is /dev/tty.usbmodem575E0031751.
-    >>> Reconnect the usb cable.
-    ```
-
-    Example of usage for 1 motor connected to the bus:
-    ```python
-    motor_name = "gripper"
-    motor_index = 6
-    motor_model = "sts3215"
-
-    config = FeetechMotorsBusConfig(
-        port="/dev/tty.usbmodem575E0031751",
-        motors={motor_name: (motor_index, motor_model)},
-    )
-    motors_bus = FeetechMotorsBus(config)
-    motors_bus.connect()
-
-    position = motors_bus.read("Present_Position")
-
-    # move from a few motor steps as an example
-    few_steps = 30
-    motors_bus.write("Goal_Position", position + few_steps)
-
-    # when done, consider disconnecting
-    motors_bus.disconnect()
-    ```
-    """
-
-    def __init__(
-        self,
-        config: FeetechMotorsBusConfig,
-    ):
-        self.port = config.port
-        self.motors = config.motors
-        self.mock = config.mock
-
-        self.model_ctrl_table = deepcopy(MODEL_CONTROL_TABLE)
-        self.model_resolution = deepcopy(MODEL_RESOLUTION)
-
-        self.port_handler = None
-        self.packet_handler = None
-        self.calibration = None
-        self.is_connected = False
-        self.group_readers = {}
-        self.group_writers = {}
-        self.logs = {}
-
-        self.track_positions = {}
-
-    def connect(self):
-        if self.is_connected:
-            raise RobotDeviceAlreadyConnectedError(
-                f"FeetechMotorsBus({self.port}) is already connected. Do not call `motors_bus.connect()` twice."
-            )
-
-        if self.mock:
-            import tests.motors.mock_scservo_sdk as scs
-        else:
-            import scservo_sdk as scs
-
-        self.port_handler = scs.PortHandler(self.port)
-        self.packet_handler = scs.PacketHandler(PROTOCOL_VERSION)
-
-        try:
-            if not self.port_handler.openPort():
-                raise OSError(f"Failed to open port '{self.port}'.")
-        except Exception:
-            traceback.print_exc()
-            print(
-                "\nTry running `python lerobot/scripts/find_motors_bus_port.py` to make sure you are using the correct port.\n"
-            )
-            raise
-
-        # Allow to read and write
-        self.is_connected = True
-
-        self.port_handler.setPacketTimeoutMillis(TIMEOUT_MS)
-
-    def reconnect(self):
-        if self.mock:
-            import tests.motors.mock_scservo_sdk as scs
-        else:
-            import scservo_sdk as scs
-
-        self.port_handler = scs.PortHandler(self.port)
-        self.packet_handler = scs.PacketHandler(PROTOCOL_VERSION)
-
-        if not self.port_handler.openPort():
-            raise OSError(f"Failed to open port '{self.port}'.")
-
-        self.is_connected = True
-
-    def are_motors_configured(self):
-        # Only check the motor indices and not baudrate, since if the motor baudrates are incorrect,
-        # a ConnectionError will be raised anyway.
-        try:
-            return (self.motor_indices == self.read("ID")).all()
-        except ConnectionError as e:
-            print(e)
-            return False
-
-    def find_motor_indices(self, possible_ids=None, num_retry=2):
-        if possible_ids is None:
-            possible_ids = range(MAX_ID_RANGE)
-
-        indices = []
-        for idx in tqdm.tqdm(possible_ids):
-            try:
-                present_idx = self.read_with_motor_ids(self.motor_models, [idx], "ID", num_retry=num_retry)[0]
-            except ConnectionError:
-                continue
-
-            if idx != present_idx:
-                # sanity check
-                raise OSError(
-                    "Motor index used to communicate through the bus is not the same as the one present in the motor memory. The motor memory might be damaged."
-                )
-            indices.append(idx)
-
-        return indices
-
-    def set_bus_baudrate(self, baudrate):
-        present_bus_baudrate = self.port_handler.getBaudRate()
-        if present_bus_baudrate != baudrate:
-            print(f"Setting bus baud rate to {baudrate}. Previously {present_bus_baudrate}.")
-            self.port_handler.setBaudRate(baudrate)
-
-            if self.port_handler.getBaudRate() != baudrate:
-                raise OSError("Failed to write bus baud rate.")
-
-    @property
-    def motor_names(self) -> list[str]:
-        return list(self.motors.keys())
-
-    @property
-    def motor_models(self) -> list[str]:
-        return [model for _, model in self.motors.values()]
-
-    @property
-    def motor_indices(self) -> list[int]:
-        return [idx for idx, _ in self.motors.values()]
-
-    def set_calibration(self, calibration: dict[str, list]):
-        self.calibration = calibration
-
-    def apply_calibration_autocorrect(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """This function apply the calibration, automatically detects out of range errors for motors values and attempt to correct.
-
-        For more info, see docstring of `apply_calibration` and `autocorrect_calibration`.
-        """
-        try:
-            values = self.apply_calibration(values, motor_names)
-        except JointOutOfRangeError as e:
-            print(e)
-            self.autocorrect_calibration(values, motor_names)
-            values = self.apply_calibration(values, motor_names)
-        return values
-
-    def apply_calibration(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """Convert from unsigned int32 joint position range [0, 2**32[ to the universal float32 nominal degree range ]-180.0, 180.0[ with
-        a "zero position" at 0 degree.
-
-        Note: We say "nominal degree range" since the motors can take values outside this range. For instance, 190 degrees, if the motor
-        rotate more than a half a turn from the zero position. However, most motors can't rotate more than 180 degrees and will stay in this range.
-
-        Joints values are original in [0, 2**32[ (unsigned int32). Each motor are expected to complete a full rotation
-        when given a goal position that is + or - their resolution. For instance, feetech xl330-m077 have a resolution of 4096, and
-        at any position in their original range, let's say the position 56734, they complete a full rotation clockwise by moving to 60830,
-        or anticlockwise by moving to 52638. The position in the original range is arbitrary and might change a lot between each motor.
-        To harmonize between motors of the same model, different robots, or even models of different brands, we propose to work
-        in the centered nominal degree range ]-180, 180[.
-        """
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        # Convert from unsigned int32 original range [0, 2**32] to signed float32 range
-        values = values.astype(np.float32)
-
-        for i, name in enumerate(motor_names):
-            calib_idx = self.calibration["motor_names"].index(name)
-            calib_mode = self.calibration["calib_mode"][calib_idx]
-
-            if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                drive_mode = self.calibration["drive_mode"][calib_idx]
-                homing_offset = self.calibration["homing_offset"][calib_idx]
-                _, model = self.motors[name]
-                resolution = self.model_resolution[model]
-
-                # Update direction of rotation of the motor to match between leader and follower.
-                # In fact, the motor of the leader for a given joint can be assembled in an
-                # opposite direction in term of rotation than the motor of the follower on the same joint.
-                if drive_mode:
-                    values[i] *= -1
-
-                # Convert from range [-2**31, 2**31[ to
-                # nominal range ]-resolution, resolution[ (e.g. ]-2048, 2048[)
-                values[i] += homing_offset
-
-                # Convert from range ]-resolution, resolution[ to
-                # universal float32 centered degree range ]-180, 180[
-                values[i] = values[i] / (resolution // 2) * HALF_TURN_DEGREE
-
-                if (values[i] < LOWER_BOUND_DEGREE) or (values[i] > UPPER_BOUND_DEGREE):
-                    raise JointOutOfRangeError(
-                        f"Wrong motor position range detected for {name}. "
-                        f"Expected to be in nominal range of [-{HALF_TURN_DEGREE}, {HALF_TURN_DEGREE}] degrees (a full rotation), "
-                        f"with a maximum range of [{LOWER_BOUND_DEGREE}, {UPPER_BOUND_DEGREE}] degrees to account for joints that can rotate a bit more, "
-                        f"but present value is {values[i]} degree. "
-                        "This might be due to a cable connection issue creating an artificial 360 degrees jump in motor values. "
-                        "You need to recalibrate by running: `python lerobot/scripts/control_robot.py calibrate`"
-                    )
-
-            elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                start_pos = self.calibration["start_pos"][calib_idx]
-                end_pos = self.calibration["end_pos"][calib_idx]
-
-                # Rescale the present position to a nominal range [0, 100] %,
-                # useful for joints with linear motions like Aloha gripper
-                values[i] = (values[i] - start_pos) / (end_pos - start_pos) * 100
-
-                if (values[i] < LOWER_BOUND_LINEAR) or (values[i] > UPPER_BOUND_LINEAR):
-                    raise JointOutOfRangeError(
-                        f"Wrong motor position range detected for {name}. "
-                        f"Expected to be in nominal range of [0, 100] % (a full linear translation), "
-                        f"with a maximum range of [{LOWER_BOUND_LINEAR}, {UPPER_BOUND_LINEAR}] % to account for some imprecision during calibration, "
-                        f"but present value is {values[i]} %. "
-                        "This might be due to a cable connection issue creating an artificial jump in motor values. "
-                        "You need to recalibrate by running: `python lerobot/scripts/control_robot.py calibrate`"
-                    )
-
-        return values
-
-    def autocorrect_calibration(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """This function automatically detects issues with values of motors after calibration, and correct for these issues.
-
-        Some motors might have values outside of expected maximum bounds after calibration.
-        For instance, for a joint in degree, its value can be outside [-270, 270] degrees, which is totally unexpected given
-        a nominal range of [-180, 180] degrees, which represents half a turn to the left or right starting from zero position.
-
-        Known issues:
-        #1: Motor value randomly shifts of a full turn, caused by hardware/connection errors.
-        #2: Motor internal homing offset is shifted of a full turn, caused by using default calibration (e.g Aloha).
-        #3: motor internal homing offset is shifted of less or more than a full turn, caused by using default calibration
-            or by human error during manual calibration.
-
-        Issues #1 and #2 can be solved by shifting the calibration homing offset by a full turn.
-        Issue #3 will be visually detected by user and potentially captured by the safety feature `max_relative_target`,
-        that will slow down the motor, raise an error asking to recalibrate. Manual recalibrating will solve the issue.
-
-        Note: A full turn corresponds to 360 degrees but also to 4096 steps for a motor resolution of 4096.
-        """
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        # Convert from unsigned int32 original range [0, 2**32] to signed float32 range
-        values = values.astype(np.float32)
-
-        for i, name in enumerate(motor_names):
-            calib_idx = self.calibration["motor_names"].index(name)
-            calib_mode = self.calibration["calib_mode"][calib_idx]
-
-            if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                drive_mode = self.calibration["drive_mode"][calib_idx]
-                homing_offset = self.calibration["homing_offset"][calib_idx]
-                _, model = self.motors[name]
-                resolution = self.model_resolution[model]
-
-                if drive_mode:
-                    values[i] *= -1
-
-                # Convert from initial range to range [-180, 180] degrees
-                calib_val = (values[i] + homing_offset) / (resolution // 2) * HALF_TURN_DEGREE
-                in_range = (calib_val > LOWER_BOUND_DEGREE) and (calib_val < UPPER_BOUND_DEGREE)
-
-                # Solve this inequality to find the factor to shift the range into [-180, 180] degrees
-                # values[i] = (values[i] + homing_offset + resolution * factor) / (resolution // 2) * HALF_TURN_DEGREE
-                # - HALF_TURN_DEGREE <= (values[i] + homing_offset + resolution * factor) / (resolution // 2) * HALF_TURN_DEGREE <= HALF_TURN_DEGREE
-                # (- HALF_TURN_DEGREE / HALF_TURN_DEGREE * (resolution // 2) - values[i] - homing_offset) / resolution <= factor <= (HALF_TURN_DEGREE / 180 * (resolution // 2) - values[i] - homing_offset) / resolution
-                low_factor = (
-                    -HALF_TURN_DEGREE / HALF_TURN_DEGREE * (resolution // 2) - values[i] - homing_offset
-                ) / resolution
-                upp_factor = (
-                    HALF_TURN_DEGREE / HALF_TURN_DEGREE * (resolution // 2) - values[i] - homing_offset
-                ) / resolution
-
-            elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                start_pos = self.calibration["start_pos"][calib_idx]
-                end_pos = self.calibration["end_pos"][calib_idx]
-
-                # Convert from initial range to range [0, 100] in %
-                calib_val = (values[i] - start_pos) / (end_pos - start_pos) * 100
-                in_range = (calib_val > LOWER_BOUND_LINEAR) and (calib_val < UPPER_BOUND_LINEAR)
-
-                # Solve this inequality to find the factor to shift the range into [0, 100] %
-                # values[i] = (values[i] - start_pos + resolution * factor) / (end_pos + resolution * factor - start_pos - resolution * factor) * 100
-                # values[i] = (values[i] - start_pos + resolution * factor) / (end_pos - start_pos) * 100
-                # 0 <= (values[i] - start_pos + resolution * factor) / (end_pos - start_pos) * 100 <= 100
-                # (start_pos - values[i]) / resolution <= factor <= (end_pos - values[i]) / resolution
-                low_factor = (start_pos - values[i]) / resolution
-                upp_factor = (end_pos - values[i]) / resolution
-
-            if not in_range:
-                # Get first integer between the two bounds
-                if low_factor < upp_factor:
-                    factor = math.ceil(low_factor)
-
-                    if factor > upp_factor:
-                        raise ValueError(f"No integer found between bounds [{low_factor=}, {upp_factor=}]")
-                else:
-                    factor = math.ceil(upp_factor)
-
-                    if factor > low_factor:
-                        raise ValueError(f"No integer found between bounds [{low_factor=}, {upp_factor=}]")
-
-                if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                    out_of_range_str = f"{LOWER_BOUND_DEGREE} < {calib_val} < {UPPER_BOUND_DEGREE} degrees"
-                    in_range_str = f"{LOWER_BOUND_DEGREE} < {calib_val} < {UPPER_BOUND_DEGREE} degrees"
-                elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                    out_of_range_str = f"{LOWER_BOUND_LINEAR} < {calib_val} < {UPPER_BOUND_LINEAR} %"
-                    in_range_str = f"{LOWER_BOUND_LINEAR} < {calib_val} < {UPPER_BOUND_LINEAR} %"
-
-                logging.warning(
-                    f"Auto-correct calibration of motor '{name}' by shifting value by {abs(factor)} full turns, "
-                    f"from '{out_of_range_str}' to '{in_range_str}'."
-                )
-
-                # A full turn corresponds to 360 degrees but also to 4096 steps for a motor resolution of 4096.
-                self.calibration["homing_offset"][calib_idx] += resolution * factor
-
-    def revert_calibration(self, values: np.ndarray | list, motor_names: list[str] | None):
-        """Inverse of `apply_calibration`."""
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        for i, name in enumerate(motor_names):
-            calib_idx = self.calibration["motor_names"].index(name)
-            calib_mode = self.calibration["calib_mode"][calib_idx]
-
-            if CalibrationMode[calib_mode] == CalibrationMode.DEGREE:
-                drive_mode = self.calibration["drive_mode"][calib_idx]
-                homing_offset = self.calibration["homing_offset"][calib_idx]
-                _, model = self.motors[name]
-                resolution = self.model_resolution[model]
-
-                # Convert from nominal 0-centered degree range [-180, 180] to
-                # 0-centered resolution range (e.g. [-2048, 2048] for resolution=4096)
-                values[i] = values[i] / HALF_TURN_DEGREE * (resolution // 2)
-
-                # Subtract the homing offsets to come back to actual motor range of values
-                # which can be arbitrary.
-                values[i] -= homing_offset
-
-                # Remove drive mode, which is the rotation direction of the motor, to come back to
-                # actual motor rotation direction which can be arbitrary.
-                if drive_mode:
-                    values[i] *= -1
-
-            elif CalibrationMode[calib_mode] == CalibrationMode.LINEAR:
-                start_pos = self.calibration["start_pos"][calib_idx]
-                end_pos = self.calibration["end_pos"][calib_idx]
-
-                # Convert from nominal lnear range of [0, 100] % to
-                # actual motor range of values which can be arbitrary.
-                values[i] = values[i] / 100 * (end_pos - start_pos) + start_pos
-
-        values = np.round(values).astype(np.int32)
-        return values
-
-    def avoid_rotation_reset(self, values, motor_names, data_name):
-        if data_name not in self.track_positions:
-            self.track_positions[data_name] = {
-                "prev": [None] * len(self.motor_names),
-                # Assume False at initialization
-                "below_zero": [False] * len(self.motor_names),
-                "above_max": [False] * len(self.motor_names),
-            }
-
-        track = self.track_positions[data_name]
-
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        for i, name in enumerate(motor_names):
-            idx = self.motor_names.index(name)
-
-            if track["prev"][idx] is None:
-                track["prev"][idx] = values[i]
-                continue
-
-            # Detect a full rotation occurred
-            if abs(track["prev"][idx] - values[i]) > 2048:
-                # Position went below 0 and got reset to 4095
-                if track["prev"][idx] < values[i]:
-                    # So we set negative value by adding a full rotation
-                    values[i] -= 4096
-
-                # Position went above 4095 and got reset to 0
-                elif track["prev"][idx] > values[i]:
-                    # So we add a full rotation
-                    values[i] += 4096
-
-            track["prev"][idx] = values[i]
-
-        return values
-
-    def read_with_motor_ids(self, motor_models, motor_ids, data_name, num_retry=NUM_READ_RETRY):
-        if self.mock:
-            import tests.motors.mock_scservo_sdk as scs
-        else:
-            import scservo_sdk as scs
-
-        return_list = True
-        if not isinstance(motor_ids, list):
-            return_list = False
-            motor_ids = [motor_ids]
-
-        assert_same_address(self.model_ctrl_table, self.motor_models, data_name)
-        addr, bytes = self.model_ctrl_table[motor_models[0]][data_name]
-        group = scs.GroupSyncRead(self.port_handler, self.packet_handler, addr, bytes)
-        for idx in motor_ids:
-            group.addParam(idx)
-
-        for _ in range(num_retry):
-            comm = group.txRxPacket()
-            if comm == scs.COMM_SUCCESS:
-                break
-
-        if comm != scs.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Read failed due to communication error on port {self.port_handler.port_name} for indices {motor_ids}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-        values = []
-        for idx in motor_ids:
-            value = group.getData(idx, addr, bytes)
-            values.append(value)
-
-        if return_list:
-            return values
-        else:
-            return values[0]
-
-    def read(self, data_name, motor_names: str | list[str] | None = None):
-        if self.mock:
-            import tests.motors.mock_scservo_sdk as scs
-        else:
-            import scservo_sdk as scs
-
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"FeetechMotorsBus({self.port}) is not connected. You need to run `motors_bus.connect()`."
-            )
-
-        start_time = time.perf_counter()
-
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        if isinstance(motor_names, str):
-            motor_names = [motor_names]
-
-        motor_ids = []
-        models = []
-        for name in motor_names:
-            motor_idx, model = self.motors[name]
-            motor_ids.append(motor_idx)
-            models.append(model)
-
-        assert_same_address(self.model_ctrl_table, models, data_name)
-        addr, bytes = self.model_ctrl_table[model][data_name]
-        group_key = get_group_sync_key(data_name, motor_names)
-
-        if data_name not in self.group_readers:
-            # Very Important to flush the buffer!
-            self.port_handler.ser.reset_output_buffer()
-            self.port_handler.ser.reset_input_buffer()
-
-            # create new group reader
-            self.group_readers[group_key] = scs.GroupSyncRead(
-                self.port_handler, self.packet_handler, addr, bytes
-            )
-            for idx in motor_ids:
-                self.group_readers[group_key].addParam(idx)
-
-        for _ in range(NUM_READ_RETRY):
-            comm = self.group_readers[group_key].txRxPacket()
-            if comm == scs.COMM_SUCCESS:
-                break
-
-        if comm != scs.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Read failed due to communication error on port {self.port} for group_key {group_key}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-        values = []
-        for idx in motor_ids:
-            value = self.group_readers[group_key].getData(idx, addr, bytes)
-            values.append(value)
-
-        values = np.array(values)
-
-        # Convert to signed int to use range [-2048, 2048] for our motor positions.
-        if data_name in CONVERT_UINT32_TO_INT32_REQUIRED:
-            values = values.astype(np.int32)
-
-        if data_name in CALIBRATION_REQUIRED:
-            values = self.avoid_rotation_reset(values, motor_names, data_name)
-
-        if data_name in CALIBRATION_REQUIRED and self.calibration is not None:
-            values = self.apply_calibration_autocorrect(values, motor_names)
-
-        # log the number of seconds it took to read the data from the motors
-        delta_ts_name = get_log_name("delta_timestamp_s", "read", data_name, motor_names)
-        self.logs[delta_ts_name] = time.perf_counter() - start_time
-
-        # log the utc time at which the data was received
-        ts_utc_name = get_log_name("timestamp_utc", "read", data_name, motor_names)
-        self.logs[ts_utc_name] = capture_timestamp_utc()
-
-        return values
-
-    def write_with_motor_ids(self, motor_models, motor_ids, data_name, values, num_retry=NUM_WRITE_RETRY):
-        if self.mock:
-            import tests.motors.mock_scservo_sdk as scs
-        else:
-            import scservo_sdk as scs
-
-        if not isinstance(motor_ids, list):
-            motor_ids = [motor_ids]
-        if not isinstance(values, list):
-            values = [values]
-
-        assert_same_address(self.model_ctrl_table, motor_models, data_name)
-        addr, bytes = self.model_ctrl_table[motor_models[0]][data_name]
-        group = scs.GroupSyncWrite(self.port_handler, self.packet_handler, addr, bytes)
-        for idx, value in zip(motor_ids, values, strict=True):
-            data = convert_to_bytes(value, bytes, self.mock)
-            group.addParam(idx, data)
-
-        for _ in range(num_retry):
-            comm = group.txPacket()
-            if comm == scs.COMM_SUCCESS:
-                break
-
-        if comm != scs.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Write failed due to communication error on port {self.port_handler.port_name} for indices {motor_ids}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-    def write(self, data_name, values: int | float | np.ndarray, motor_names: str | list[str] | None = None):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"FeetechMotorsBus({self.port}) is not connected. You need to run `motors_bus.connect()`."
-            )
-
-        start_time = time.perf_counter()
-
-        if self.mock:
-            import tests.motors.mock_scservo_sdk as scs
-        else:
-            import scservo_sdk as scs
-
-        if motor_names is None:
-            motor_names = self.motor_names
-
-        if isinstance(motor_names, str):
-            motor_names = [motor_names]
-
-        if isinstance(values, (int, float, np.integer)):
-            values = [int(values)] * len(motor_names)
-
-        values = np.array(values)
-
-        motor_ids = []
-        models = []
-        for name in motor_names:
-            motor_idx, model = self.motors[name]
-            motor_ids.append(motor_idx)
-            models.append(model)
-
-        if data_name in CALIBRATION_REQUIRED and self.calibration is not None:
-            values = self.revert_calibration(values, motor_names)
-
-        values = values.tolist()
-
-        assert_same_address(self.model_ctrl_table, models, data_name)
-        addr, bytes = self.model_ctrl_table[model][data_name]
-        group_key = get_group_sync_key(data_name, motor_names)
-
-        init_group = data_name not in self.group_readers
-        if init_group:
-            self.group_writers[group_key] = scs.GroupSyncWrite(
-                self.port_handler, self.packet_handler, addr, bytes
-            )
-
-        for idx, value in zip(motor_ids, values, strict=True):
-            data = convert_to_bytes(value, bytes, self.mock)
-            if init_group:
-                self.group_writers[group_key].addParam(idx, data)
-            else:
-                self.group_writers[group_key].changeParam(idx, data)
-
-        comm = self.group_writers[group_key].txPacket()
-        if comm != scs.COMM_SUCCESS:
-            raise ConnectionError(
-                f"Write failed due to communication error on port {self.port} for group_key {group_key}: "
-                f"{self.packet_handler.getTxRxResult(comm)}"
-            )
-
-        # log the number of seconds it took to write the data to the motors
-        delta_ts_name = get_log_name("delta_timestamp_s", "write", data_name, motor_names)
-        self.logs[delta_ts_name] = time.perf_counter() - start_time
-
-        # TODO(rcadene): should we log the time before sending the write command?
-        # log the utc time when the write has been completed
-        ts_utc_name = get_log_name("timestamp_utc", "write", data_name, motor_names)
-        self.logs[ts_utc_name] = capture_timestamp_utc()
-
-    def disconnect(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                f"FeetechMotorsBus({self.port}) is not connected. Try running `motors_bus.connect()` first."
-            )
-
-        if self.port_handler is not None:
-            self.port_handler.closePort()
-            self.port_handler = None
-
-        self.packet_handler = None
-        self.group_readers = {}
-        self.group_writers = {}
-        self.is_connected = False
-
-    def __del__(self):
-        if getattr(self, "is_connected", False):
-            self.disconnect()

lerobot/common/robot_devices/motors/utils.py

@@ -1,67 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from typing import Protocol
-
-from lerobot.common.robot_devices.motors.configs import (
-    DynamixelMotorsBusConfig,
-    FeetechMotorsBusConfig,
-    MotorsBusConfig,
-)
-
-
-class MotorsBus(Protocol):
-    def motor_names(self): ...
-    def set_calibration(self): ...
-    def apply_calibration(self): ...
-    def revert_calibration(self): ...
-    def read(self): ...
-    def write(self): ...
-
-
-def make_motors_buses_from_configs(motors_bus_configs: dict[str, MotorsBusConfig]) -> list[MotorsBus]:
-    motors_buses = {}
-
-    for key, cfg in motors_bus_configs.items():
-        if cfg.type == "dynamixel":
-            from lerobot.common.robot_devices.motors.dynamixel import DynamixelMotorsBus
-
-            motors_buses[key] = DynamixelMotorsBus(cfg)
-
-        elif cfg.type == "feetech":
-            from lerobot.common.robot_devices.motors.feetech import FeetechMotorsBus
-
-            motors_buses[key] = FeetechMotorsBus(cfg)
-
-        else:
-            raise ValueError(f"The motor type '{cfg.type}' is not valid.")
-
-    return motors_buses
-
-
-def make_motors_bus(motor_type: str, **kwargs) -> MotorsBus:
-    if motor_type == "dynamixel":
-        from lerobot.common.robot_devices.motors.dynamixel import DynamixelMotorsBus
-
-        config = DynamixelMotorsBusConfig(**kwargs)
-        return DynamixelMotorsBus(config)
-
-    elif motor_type == "feetech":
-        from lerobot.common.robot_devices.motors.feetech import FeetechMotorsBus
-
-        config = FeetechMotorsBusConfig(**kwargs)
-        return FeetechMotorsBus(config)
-
-    else:
-        raise ValueError(f"The motor type '{motor_type}' is not valid.")

lerobot/common/robot_devices/robots/configs.py

@@ -1,676 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import abc
-from dataclasses import dataclass, field
-from typing import Sequence
-
-import draccus
-
-from lerobot.common.robot_devices.cameras.configs import (
-    CameraConfig,
-    IntelRealSenseCameraConfig,
-    OpenCVCameraConfig,
-)
-from lerobot.common.robot_devices.motors.configs import (
-    DynamixelMotorsBusConfig,
-    FeetechMotorsBusConfig,
-    MotorsBusConfig,
-)
-
-
-@dataclass
-class RobotConfig(draccus.ChoiceRegistry, abc.ABC):
-    @property
-    def type(self) -> str:
-        return self.get_choice_name(self.__class__)
-
-
-# TODO(rcadene, aliberts): remove ManipulatorRobotConfig abstraction
-@dataclass
-class ManipulatorRobotConfig(RobotConfig):
-    leader_arms: dict[str, MotorsBusConfig] = field(default_factory=lambda: {})
-    follower_arms: dict[str, MotorsBusConfig] = field(default_factory=lambda: {})
-    cameras: dict[str, CameraConfig] = field(default_factory=lambda: {})
-
-    # Optionally limit the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length
-    # as the number of motors in your follower arms (assumes all follower arms have the same number of
-    # motors).
-    max_relative_target: list[float] | float | None = None
-
-    # Optionally set the leader arm in torque mode with the gripper motor set to this angle. This makes it
-    # possible to squeeze the gripper and have it spring back to an open position on its own. If None, the
-    # gripper is not put in torque mode.
-    gripper_open_degree: float | None = None
-
-    mock: bool = False
-
-    def __post_init__(self):
-        if self.mock:
-            for arm in self.leader_arms.values():
-                if not arm.mock:
-                    arm.mock = True
-            for arm in self.follower_arms.values():
-                if not arm.mock:
-                    arm.mock = True
-            for cam in self.cameras.values():
-                if not cam.mock:
-                    cam.mock = True
-
-        if self.max_relative_target is not None and isinstance(self.max_relative_target, Sequence):
-            for name in self.follower_arms:
-                if len(self.follower_arms[name].motors) != len(self.max_relative_target):
-                    raise ValueError(
-                        f"len(max_relative_target)={len(self.max_relative_target)} but the follower arm with name {name} has "
-                        f"{len(self.follower_arms[name].motors)} motors. Please make sure that the "
-                        f"`max_relative_target` list has as many parameters as there are motors per arm. "
-                        "Note: This feature does not yet work with robots where different follower arms have "
-                        "different numbers of motors."
-                    )
-
-
-@RobotConfig.register_subclass("aloha")
-@dataclass
-class AlohaRobotConfig(ManipulatorRobotConfig):
-    # Specific to Aloha, LeRobot comes with default calibration files. Assuming the motors have been
-    # properly assembled, no manual calibration step is expected. If you need to run manual calibration,
-    # simply update this path to ".cache/calibration/aloha"
-    calibration_dir: str = ".cache/calibration/aloha_default"
-
-    # /!\ FOR SAFETY, READ THIS /!\
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    # For Aloha, for every goal position request, motor rotations are capped at 5 degrees by default.
-    # When you feel more confident with teleoperation or running the policy, you can extend
-    # this safety limit and even removing it by setting it to `null`.
-    # Also, everything is expected to work safely out-of-the-box, but we highly advise to
-    # first try to teleoperate the grippers only (by commenting out the rest of the motors in this yaml),
-    # then to gradually add more motors (by uncommenting), until you can teleoperate both arms fully
-    max_relative_target: int | None = 5
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "left": DynamixelMotorsBusConfig(
-                # window_x
-                port="/dev/ttyDXL_leader_left",
-                motors={
-                    # name: (index, model)
-                    "waist": [1, "xm430-w350"],
-                    "shoulder": [2, "xm430-w350"],
-                    "shoulder_shadow": [3, "xm430-w350"],
-                    "elbow": [4, "xm430-w350"],
-                    "elbow_shadow": [5, "xm430-w350"],
-                    "forearm_roll": [6, "xm430-w350"],
-                    "wrist_angle": [7, "xm430-w350"],
-                    "wrist_rotate": [8, "xl430-w250"],
-                    "gripper": [9, "xc430-w150"],
-                },
-            ),
-            "right": DynamixelMotorsBusConfig(
-                # window_x
-                port="/dev/ttyDXL_leader_right",
-                motors={
-                    # name: (index, model)
-                    "waist": [1, "xm430-w350"],
-                    "shoulder": [2, "xm430-w350"],
-                    "shoulder_shadow": [3, "xm430-w350"],
-                    "elbow": [4, "xm430-w350"],
-                    "elbow_shadow": [5, "xm430-w350"],
-                    "forearm_roll": [6, "xm430-w350"],
-                    "wrist_angle": [7, "xm430-w350"],
-                    "wrist_rotate": [8, "xl430-w250"],
-                    "gripper": [9, "xc430-w150"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "left": DynamixelMotorsBusConfig(
-                port="/dev/ttyDXL_follower_left",
-                motors={
-                    # name: (index, model)
-                    "waist": [1, "xm540-w270"],
-                    "shoulder": [2, "xm540-w270"],
-                    "shoulder_shadow": [3, "xm540-w270"],
-                    "elbow": [4, "xm540-w270"],
-                    "elbow_shadow": [5, "xm540-w270"],
-                    "forearm_roll": [6, "xm540-w270"],
-                    "wrist_angle": [7, "xm540-w270"],
-                    "wrist_rotate": [8, "xm430-w350"],
-                    "gripper": [9, "xm430-w350"],
-                },
-            ),
-            "right": DynamixelMotorsBusConfig(
-                port="/dev/ttyDXL_follower_right",
-                motors={
-                    # name: (index, model)
-                    "waist": [1, "xm540-w270"],
-                    "shoulder": [2, "xm540-w270"],
-                    "shoulder_shadow": [3, "xm540-w270"],
-                    "elbow": [4, "xm540-w270"],
-                    "elbow_shadow": [5, "xm540-w270"],
-                    "forearm_roll": [6, "xm540-w270"],
-                    "wrist_angle": [7, "xm540-w270"],
-                    "wrist_rotate": [8, "xm430-w350"],
-                    "gripper": [9, "xm430-w350"],
-                },
-            ),
-        }
-    )
-
-    # Troubleshooting: If one of your IntelRealSense cameras freeze during
-    # data recording due to bandwidth limit, you might need to plug the camera
-    # on another USB hub or PCIe card.
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "cam_high": IntelRealSenseCameraConfig(
-                serial_number=128422271347,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "cam_low": IntelRealSenseCameraConfig(
-                serial_number=130322270656,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "cam_left_wrist": IntelRealSenseCameraConfig(
-                serial_number=218622272670,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "cam_right_wrist": IntelRealSenseCameraConfig(
-                serial_number=130322272300,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("koch")
-@dataclass
-class KochRobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/koch"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0085511",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl330-m077"],
-                    "shoulder_lift": [2, "xl330-m077"],
-                    "elbow_flex": [3, "xl330-m077"],
-                    "wrist_flex": [4, "xl330-m077"],
-                    "wrist_roll": [5, "xl330-m077"],
-                    "gripper": [6, "xl330-m077"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl430-w250"],
-                    "shoulder_lift": [2, "xl430-w250"],
-                    "elbow_flex": [3, "xl330-m288"],
-                    "wrist_flex": [4, "xl330-m288"],
-                    "wrist_roll": [5, "xl330-m288"],
-                    "gripper": [6, "xl330-m288"],
-                },
-            ),
-        }
-    )
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "laptop": OpenCVCameraConfig(
-                camera_index=0,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "phone": OpenCVCameraConfig(
-                camera_index=1,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    # ~ Koch specific settings ~
-    # Sets the leader arm in torque mode with the gripper motor set to this angle. This makes it possible
-    # to squeeze the gripper and have it spring back to an open position on its own.
-    gripper_open_degree: float = 35.156
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("koch_bimanual")
-@dataclass
-class KochBimanualRobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/koch_bimanual"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "left": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0085511",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl330-m077"],
-                    "shoulder_lift": [2, "xl330-m077"],
-                    "elbow_flex": [3, "xl330-m077"],
-                    "wrist_flex": [4, "xl330-m077"],
-                    "wrist_roll": [5, "xl330-m077"],
-                    "gripper": [6, "xl330-m077"],
-                },
-            ),
-            "right": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem575E0031751",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl330-m077"],
-                    "shoulder_lift": [2, "xl330-m077"],
-                    "elbow_flex": [3, "xl330-m077"],
-                    "wrist_flex": [4, "xl330-m077"],
-                    "wrist_roll": [5, "xl330-m077"],
-                    "gripper": [6, "xl330-m077"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "left": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl430-w250"],
-                    "shoulder_lift": [2, "xl430-w250"],
-                    "elbow_flex": [3, "xl330-m288"],
-                    "wrist_flex": [4, "xl330-m288"],
-                    "wrist_roll": [5, "xl330-m288"],
-                    "gripper": [6, "xl330-m288"],
-                },
-            ),
-            "right": DynamixelMotorsBusConfig(
-                port="/dev/tty.usbmodem575E0032081",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "xl430-w250"],
-                    "shoulder_lift": [2, "xl430-w250"],
-                    "elbow_flex": [3, "xl330-m288"],
-                    "wrist_flex": [4, "xl330-m288"],
-                    "wrist_roll": [5, "xl330-m288"],
-                    "gripper": [6, "xl330-m288"],
-                },
-            ),
-        }
-    )
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "laptop": OpenCVCameraConfig(
-                camera_index=0,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "phone": OpenCVCameraConfig(
-                camera_index=1,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    # ~ Koch specific settings ~
-    # Sets the leader arm in torque mode with the gripper motor set to this angle. This makes it possible
-    # to squeeze the gripper and have it spring back to an open position on its own.
-    gripper_open_degree: float = 35.156
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("moss")
-@dataclass
-class MossRobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/moss"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "laptop": OpenCVCameraConfig(
-                camera_index=0,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "phone": OpenCVCameraConfig(
-                camera_index=1,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("so101")
-@dataclass
-class So101RobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/so101"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "laptop": OpenCVCameraConfig(
-                camera_index=0,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "phone": OpenCVCameraConfig(
-                camera_index=1,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("so100")
-@dataclass
-class So100RobotConfig(ManipulatorRobotConfig):
-    calibration_dir: str = ".cache/calibration/so100"
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem58760431091",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0076891",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "laptop": OpenCVCameraConfig(
-                camera_index=0,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-            "phone": OpenCVCameraConfig(
-                camera_index=1,
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("stretch")
-@dataclass
-class StretchRobotConfig(RobotConfig):
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "navigation": OpenCVCameraConfig(
-                camera_index="/dev/hello-nav-head-camera",
-                fps=10,
-                width=1280,
-                height=720,
-                rotation=-90,
-            ),
-            "head": IntelRealSenseCameraConfig(
-                name="Intel RealSense D435I",
-                fps=30,
-                width=640,
-                height=480,
-                rotation=90,
-            ),
-            "wrist": IntelRealSenseCameraConfig(
-                name="Intel RealSense D405",
-                fps=30,
-                width=640,
-                height=480,
-            ),
-        }
-    )
-
-    mock: bool = False
-
-
-@RobotConfig.register_subclass("lekiwi")
-@dataclass
-class LeKiwiRobotConfig(RobotConfig):
-    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
-    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
-    # the number of motors in your follower arms.
-    max_relative_target: int | None = None
-
-    # Network Configuration
-    ip: str = "192.168.0.193"
-    port: int = 5555
-    video_port: int = 5556
-
-    cameras: dict[str, CameraConfig] = field(
-        default_factory=lambda: {
-            "front": OpenCVCameraConfig(
-                camera_index="/dev/video0", fps=30, width=640, height=480, rotation=90
-            ),
-            "wrist": OpenCVCameraConfig(
-                camera_index="/dev/video2", fps=30, width=640, height=480, rotation=180
-            ),
-        }
-    )
-
-    calibration_dir: str = ".cache/calibration/lekiwi"
-
-    leader_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/tty.usbmodem585A0077581",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                },
-            ),
-        }
-    )
-
-    follower_arms: dict[str, MotorsBusConfig] = field(
-        default_factory=lambda: {
-            "main": FeetechMotorsBusConfig(
-                port="/dev/ttyACM0",
-                motors={
-                    # name: (index, model)
-                    "shoulder_pan": [1, "sts3215"],
-                    "shoulder_lift": [2, "sts3215"],
-                    "elbow_flex": [3, "sts3215"],
-                    "wrist_flex": [4, "sts3215"],
-                    "wrist_roll": [5, "sts3215"],
-                    "gripper": [6, "sts3215"],
-                    "left_wheel": (7, "sts3215"),
-                    "back_wheel": (8, "sts3215"),
-                    "right_wheel": (9, "sts3215"),
-                },
-            ),
-        }
-    )
-
-    teleop_keys: dict[str, str] = field(
-        default_factory=lambda: {
-            # Movement
-            "forward": "w",
-            "backward": "s",
-            "left": "a",
-            "right": "d",
-            "rotate_left": "z",
-            "rotate_right": "x",
-            # Speed control
-            "speed_up": "r",
-            "speed_down": "f",
-            # quit teleop
-            "quit": "q",
-        }
-    )
-
-    mock: bool = False

lerobot/common/robot_devices/robots/dynamixel_calibration.py

@@ -1,144 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-"""Logic to calibrate a robot arm built with dynamixel motors"""
-# TODO(rcadene, aliberts): move this logic into the robot code when refactoring
-
-import numpy as np
-
-from lerobot.common.robot_devices.motors.dynamixel import (
-    CalibrationMode,
-    TorqueMode,
-    convert_degrees_to_steps,
-)
-from lerobot.common.robot_devices.motors.utils import MotorsBus
-
-URL_TEMPLATE = (
-    "https://raw.githubusercontent.com/huggingface/lerobot/main/media/{robot}/{arm}_{position}.webp"
-)
-
-# The following positions are provided in nominal degree range ]-180, +180[
-# For more info on these constants, see comments in the code where they get used.
-ZERO_POSITION_DEGREE = 0
-ROTATED_POSITION_DEGREE = 90
-
-
-def assert_drive_mode(drive_mode):
-    # `drive_mode` is in [0,1] with 0 means original rotation direction for the motor, and 1 means inverted.
-    if not np.all(np.isin(drive_mode, [0, 1])):
-        raise ValueError(f"`drive_mode` contains values other than 0 or 1: ({drive_mode})")
-
-
-def apply_drive_mode(position, drive_mode):
-    assert_drive_mode(drive_mode)
-    # Convert `drive_mode` from [0, 1] with 0 indicates original rotation direction and 1 inverted,
-    # to [-1, 1] with 1 indicates original rotation direction and -1 inverted.
-    signed_drive_mode = -(drive_mode * 2 - 1)
-    position *= signed_drive_mode
-    return position
-
-
-def compute_nearest_rounded_position(position, models):
-    delta_turn = convert_degrees_to_steps(ROTATED_POSITION_DEGREE, models)
-    nearest_pos = np.round(position.astype(float) / delta_turn) * delta_turn
-    return nearest_pos.astype(position.dtype)
-
-
-def run_arm_calibration(arm: MotorsBus, robot_type: str, arm_name: str, arm_type: str):
-    """This function ensures that a neural network trained on data collected on a given robot
-    can work on another robot. For instance before calibration, setting a same goal position
-    for each motor of two different robots will get two very different positions. But after calibration,
-    the two robots will move to the same position.To this end, this function computes the homing offset
-    and the drive mode for each motor of a given robot.
-
-    Homing offset is used to shift the motor position to a ]-2048, +2048[ nominal range (when the motor uses 2048 steps
-    to complete a half a turn). This range is set around an arbitrary "zero position" corresponding to all motor positions
-    being 0. During the calibration process, you will need to manually move the robot to this "zero position".
-
-    Drive mode is used to invert the rotation direction of the motor. This is useful when some motors have been assembled
-    in the opposite orientation for some robots. During the calibration process, you will need to manually move the robot
-    to the "rotated position".
-
-    After calibration, the homing offsets and drive modes are stored in a cache.
-
-    Example of usage:
-    ```python
-    run_arm_calibration(arm, "koch", "left", "follower")
-    ```
-    """
-    if (arm.read("Torque_Enable") != TorqueMode.DISABLED.value).any():
-        raise ValueError("To run calibration, the torque must be disabled on all motors.")
-
-    print(f"\nRunning calibration of {robot_type} {arm_name} {arm_type}...")
-
-    print("\nMove arm to zero position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="zero"))
-    input("Press Enter to continue...")
-
-    # We arbitrarily chose our zero target position to be a straight horizontal position with gripper upwards and closed.
-    # It is easy to identify and all motors are in a "quarter turn" position. Once calibration is done, this position will
-    # correspond to every motor angle being 0. If you set all 0 as Goal Position, the arm will move in this position.
-    zero_target_pos = convert_degrees_to_steps(ZERO_POSITION_DEGREE, arm.motor_models)
-
-    # Compute homing offset so that `present_position + homing_offset ~= target_position`.
-    zero_pos = arm.read("Present_Position")
-    zero_nearest_pos = compute_nearest_rounded_position(zero_pos, arm.motor_models)
-    homing_offset = zero_target_pos - zero_nearest_pos
-
-    # The rotated target position corresponds to a rotation of a quarter turn from the zero position.
-    # This allows to identify the rotation direction of each motor.
-    # For instance, if the motor rotates 90 degree, and its value is -90 after applying the homing offset, then we know its rotation direction
-    # is inverted. However, for the calibration being successful, we need everyone to follow the same target position.
-    # Sometimes, there is only one possible rotation direction. For instance, if the gripper is closed, there is only one direction which
-    # corresponds to opening the gripper. When the rotation direction is ambiguous, we arbitrarily rotate clockwise from the point of view
-    # of the previous motor in the kinetic chain.
-    print("\nMove arm to rotated target position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="rotated"))
-    input("Press Enter to continue...")
-
-    rotated_target_pos = convert_degrees_to_steps(ROTATED_POSITION_DEGREE, arm.motor_models)
-
-    # Find drive mode by rotating each motor by a quarter of a turn.
-    # Drive mode indicates if the motor rotation direction should be inverted (=1) or not (=0).
-    rotated_pos = arm.read("Present_Position")
-    drive_mode = (rotated_pos < zero_pos).astype(np.int32)
-
-    # Re-compute homing offset to take into account drive mode
-    rotated_drived_pos = apply_drive_mode(rotated_pos, drive_mode)
-    rotated_nearest_pos = compute_nearest_rounded_position(rotated_drived_pos, arm.motor_models)
-    homing_offset = rotated_target_pos - rotated_nearest_pos
-
-    print("\nMove arm to rest position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="rest"))
-    input("Press Enter to continue...")
-    print()
-
-    # Joints with rotational motions are expressed in degrees in nominal range of [-180, 180]
-    calib_mode = [CalibrationMode.DEGREE.name] * len(arm.motor_names)
-
-    # TODO(rcadene): make type of joints (DEGREE or LINEAR) configurable from yaml?
-    if robot_type in ["aloha"] and "gripper" in arm.motor_names:
-        # Joints with linear motions (like gripper of Aloha) are expressed in nominal range of [0, 100]
-        calib_idx = arm.motor_names.index("gripper")
-        calib_mode[calib_idx] = CalibrationMode.LINEAR.name
-
-    calib_data = {
-        "homing_offset": homing_offset.tolist(),
-        "drive_mode": drive_mode.tolist(),
-        "start_pos": zero_pos.tolist(),
-        "end_pos": rotated_pos.tolist(),
-        "calib_mode": calib_mode,
-        "motor_names": arm.motor_names,
-    }
-    return calib_data

lerobot/common/robot_devices/robots/feetech_calibration.py

@@ -1,506 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-"""Logic to calibrate a robot arm built with feetech motors"""
-# TODO(rcadene, aliberts): move this logic into the robot code when refactoring
-
-import time
-
-import numpy as np
-
-from lerobot.common.robot_devices.motors.feetech import (
-    CalibrationMode,
-    TorqueMode,
-    convert_degrees_to_steps,
-)
-from lerobot.common.robot_devices.motors.utils import MotorsBus
-
-URL_TEMPLATE = (
-    "https://raw.githubusercontent.com/huggingface/lerobot/main/media/{robot}/{arm}_{position}.webp"
-)
-
-# The following positions are provided in nominal degree range ]-180, +180[
-# For more info on these constants, see comments in the code where they get used.
-ZERO_POSITION_DEGREE = 0
-ROTATED_POSITION_DEGREE = 90
-
-
-def reset_middle_positions(arm: MotorsBus):
-    input("Please move the robot to the new middle position for calibration, then press Enter...")
-    # Write 128 to Torque_Enable for all motors.
-    arm.write("Torque_Enable", 128)
-
-
-def assert_drive_mode(drive_mode):
-    # `drive_mode` is in [0,1] with 0 means original rotation direction for the motor, and 1 means inverted.
-    if not np.all(np.isin(drive_mode, [0, 1])):
-        raise ValueError(f"`drive_mode` contains values other than 0 or 1: ({drive_mode})")
-
-
-def apply_drive_mode(position, drive_mode):
-    assert_drive_mode(drive_mode)
-    # Convert `drive_mode` from [0, 1] with 0 indicates original rotation direction and 1 inverted,
-    # to [-1, 1] with 1 indicates original rotation direction and -1 inverted.
-    signed_drive_mode = -(drive_mode * 2 - 1)
-    position *= signed_drive_mode
-    return position
-
-
-def move_until_block(arm, motor_name, positive_direction=True, while_move_hook=None):
-    count = 0
-    while True:
-        present_pos = arm.read("Present_Position", motor_name)
-        if positive_direction:
-            # Move +100 steps every time. Lower the steps to lower the speed at which the arm moves.
-            arm.write("Goal_Position", present_pos + 100, motor_name)
-        else:
-            arm.write("Goal_Position", present_pos - 100, motor_name)
-
-        if while_move_hook is not None:
-            while_move_hook()
-
-        present_pos = arm.read("Present_Position", motor_name).item()
-        present_speed = arm.read("Present_Speed", motor_name).item()
-        present_current = arm.read("Present_Current", motor_name).item()
-        # present_load = arm.read("Present_Load", motor_name).item()
-        # present_voltage = arm.read("Present_Voltage", motor_name).item()
-        # present_temperature = arm.read("Present_Temperature", motor_name).item()
-
-        # print(f"{present_pos=}")
-        # print(f"{present_speed=}")
-        # print(f"{present_current=}")
-        # print(f"{present_load=}")
-        # print(f"{present_voltage=}")
-        # print(f"{present_temperature=}")
-
-        if present_speed == 0 and present_current > 40:
-            count += 1
-            if count > 100 or present_current > 300:
-                return present_pos
-        else:
-            count = 0
-
-
-def move_to_calibrate(
-    arm,
-    motor_name,
-    invert_drive_mode=False,
-    positive_first=True,
-    in_between_move_hook=None,
-    while_move_hook=None,
-):
-    initial_pos = arm.read("Present_Position", motor_name)
-
-    if positive_first:
-        p_present_pos = move_until_block(
-            arm, motor_name, positive_direction=True, while_move_hook=while_move_hook
-        )
-    else:
-        n_present_pos = move_until_block(
-            arm, motor_name, positive_direction=False, while_move_hook=while_move_hook
-        )
-
-    if in_between_move_hook is not None:
-        in_between_move_hook()
-
-    if positive_first:
-        n_present_pos = move_until_block(
-            arm, motor_name, positive_direction=False, while_move_hook=while_move_hook
-        )
-    else:
-        p_present_pos = move_until_block(
-            arm, motor_name, positive_direction=True, while_move_hook=while_move_hook
-        )
-
-    zero_pos = (n_present_pos + p_present_pos) / 2
-
-    calib_data = {
-        "initial_pos": initial_pos,
-        "homing_offset": zero_pos if invert_drive_mode else -zero_pos,
-        "invert_drive_mode": invert_drive_mode,
-        "drive_mode": -1 if invert_drive_mode else 0,
-        "zero_pos": zero_pos,
-        "start_pos": n_present_pos if invert_drive_mode else p_present_pos,
-        "end_pos": p_present_pos if invert_drive_mode else n_present_pos,
-    }
-    return calib_data
-
-
-def apply_offset(calib, offset):
-    calib["zero_pos"] += offset
-    if calib["drive_mode"]:
-        calib["homing_offset"] += offset
-    else:
-        calib["homing_offset"] -= offset
-    return calib
-
-
-def run_arm_auto_calibration(arm: MotorsBus, robot_type: str, arm_name: str, arm_type: str):
-    if robot_type == "so100":
-        return run_arm_auto_calibration_so100(arm, robot_type, arm_name, arm_type)
-    elif robot_type == "moss":
-        return run_arm_auto_calibration_moss(arm, robot_type, arm_name, arm_type)
-    else:
-        raise ValueError(robot_type)
-
-
-def run_arm_auto_calibration_so100(arm: MotorsBus, robot_type: str, arm_name: str, arm_type: str):
-    """All the offsets and magic numbers are hand tuned, and are unique to SO-100 follower arms"""
-    if (arm.read("Torque_Enable") != TorqueMode.DISABLED.value).any():
-        raise ValueError("To run calibration, the torque must be disabled on all motors.")
-
-    if not (robot_type == "so100" and arm_type == "follower"):
-        raise NotImplementedError("Auto calibration only supports the follower of so100 arms for now.")
-
-    print(f"\nRunning calibration of {robot_type} {arm_name} {arm_type}...")
-
-    print("\nMove arm to initial position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="initial"))
-    input("Press Enter to continue...")
-
-    # Lower the acceleration of the motors (in [0,254])
-    initial_acceleration = arm.read("Acceleration")
-    arm.write("Lock", 0)
-    arm.write("Acceleration", 10)
-    time.sleep(1)
-
-    arm.write("Torque_Enable", TorqueMode.ENABLED.value)
-
-    print(f'{arm.read("Present_Position", "elbow_flex")=}')
-
-    calib = {}
-
-    init_wf_pos = arm.read("Present_Position", "wrist_flex")
-    init_sl_pos = arm.read("Present_Position", "shoulder_lift")
-    init_ef_pos = arm.read("Present_Position", "elbow_flex")
-    arm.write("Goal_Position", init_wf_pos - 800, "wrist_flex")
-    arm.write("Goal_Position", init_sl_pos + 150 + 1024, "shoulder_lift")
-    arm.write("Goal_Position", init_ef_pos - 2048, "elbow_flex")
-    time.sleep(2)
-
-    print("Calibrate shoulder_pan")
-    calib["shoulder_pan"] = move_to_calibrate(arm, "shoulder_pan")
-    arm.write("Goal_Position", calib["shoulder_pan"]["zero_pos"], "shoulder_pan")
-    time.sleep(1)
-
-    print("Calibrate gripper")
-    calib["gripper"] = move_to_calibrate(arm, "gripper", invert_drive_mode=True)
-    time.sleep(1)
-
-    print("Calibrate wrist_flex")
-    calib["wrist_flex"] = move_to_calibrate(arm, "wrist_flex")
-    calib["wrist_flex"] = apply_offset(calib["wrist_flex"], offset=80)
-
-    def in_between_move_hook():
-        nonlocal arm, calib
-        time.sleep(2)
-        ef_pos = arm.read("Present_Position", "elbow_flex")
-        sl_pos = arm.read("Present_Position", "shoulder_lift")
-        arm.write("Goal_Position", ef_pos + 1024, "elbow_flex")
-        arm.write("Goal_Position", sl_pos - 1024, "shoulder_lift")
-        time.sleep(2)
-
-    print("Calibrate elbow_flex")
-    calib["elbow_flex"] = move_to_calibrate(
-        arm, "elbow_flex", positive_first=False, in_between_move_hook=in_between_move_hook
-    )
-    calib["elbow_flex"] = apply_offset(calib["elbow_flex"], offset=80 - 1024)
-
-    arm.write("Goal_Position", calib["elbow_flex"]["zero_pos"] + 1024 + 512, "elbow_flex")
-    time.sleep(1)
-
-    def in_between_move_hook():
-        nonlocal arm, calib
-        arm.write("Goal_Position", calib["elbow_flex"]["zero_pos"], "elbow_flex")
-
-    print("Calibrate shoulder_lift")
-    calib["shoulder_lift"] = move_to_calibrate(
-        arm,
-        "shoulder_lift",
-        invert_drive_mode=True,
-        positive_first=False,
-        in_between_move_hook=in_between_move_hook,
-    )
-    # add an 30 steps as offset to align with body
-    calib["shoulder_lift"] = apply_offset(calib["shoulder_lift"], offset=1024 - 50)
-
-    def while_move_hook():
-        nonlocal arm, calib
-        positions = {
-            "shoulder_lift": round(calib["shoulder_lift"]["zero_pos"] - 1600),
-            "elbow_flex": round(calib["elbow_flex"]["zero_pos"] + 1700),
-            "wrist_flex": round(calib["wrist_flex"]["zero_pos"] + 800),
-            "gripper": round(calib["gripper"]["end_pos"]),
-        }
-        arm.write("Goal_Position", list(positions.values()), list(positions.keys()))
-
-    arm.write("Goal_Position", round(calib["shoulder_lift"]["zero_pos"] - 1600), "shoulder_lift")
-    time.sleep(2)
-    arm.write("Goal_Position", round(calib["elbow_flex"]["zero_pos"] + 1700), "elbow_flex")
-    time.sleep(2)
-    arm.write("Goal_Position", round(calib["wrist_flex"]["zero_pos"] + 800), "wrist_flex")
-    time.sleep(2)
-    arm.write("Goal_Position", round(calib["gripper"]["end_pos"]), "gripper")
-    time.sleep(2)
-
-    print("Calibrate wrist_roll")
-    calib["wrist_roll"] = move_to_calibrate(
-        arm, "wrist_roll", invert_drive_mode=True, positive_first=False, while_move_hook=while_move_hook
-    )
-
-    arm.write("Goal_Position", calib["wrist_roll"]["zero_pos"], "wrist_roll")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["gripper"]["start_pos"], "gripper")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"], "wrist_flex")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["elbow_flex"]["zero_pos"] + 2048, "elbow_flex")
-    arm.write("Goal_Position", calib["shoulder_lift"]["zero_pos"] - 2048, "shoulder_lift")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["shoulder_pan"]["zero_pos"], "shoulder_pan")
-    time.sleep(1)
-
-    calib_modes = []
-    for name in arm.motor_names:
-        if name == "gripper":
-            calib_modes.append(CalibrationMode.LINEAR.name)
-        else:
-            calib_modes.append(CalibrationMode.DEGREE.name)
-
-    calib_dict = {
-        "homing_offset": [calib[name]["homing_offset"] for name in arm.motor_names],
-        "drive_mode": [calib[name]["drive_mode"] for name in arm.motor_names],
-        "start_pos": [calib[name]["start_pos"] for name in arm.motor_names],
-        "end_pos": [calib[name]["end_pos"] for name in arm.motor_names],
-        "calib_mode": calib_modes,
-        "motor_names": arm.motor_names,
-    }
-
-    # Re-enable original accerlation
-    arm.write("Lock", 0)
-    arm.write("Acceleration", initial_acceleration)
-    time.sleep(1)
-
-    return calib_dict
-
-
-def run_arm_auto_calibration_moss(arm: MotorsBus, robot_type: str, arm_name: str, arm_type: str):
-    """All the offsets and magic numbers are hand tuned, and are unique to SO-100 follower arms"""
-    if (arm.read("Torque_Enable") != TorqueMode.DISABLED.value).any():
-        raise ValueError("To run calibration, the torque must be disabled on all motors.")
-
-    if not (robot_type == "moss" and arm_type == "follower"):
-        raise NotImplementedError("Auto calibration only supports the follower of moss arms for now.")
-
-    print(f"\nRunning calibration of {robot_type} {arm_name} {arm_type}...")
-
-    print("\nMove arm to initial position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="initial"))
-    input("Press Enter to continue...")
-
-    # Lower the acceleration of the motors (in [0,254])
-    initial_acceleration = arm.read("Acceleration")
-    arm.write("Lock", 0)
-    arm.write("Acceleration", 10)
-    time.sleep(1)
-
-    arm.write("Torque_Enable", TorqueMode.ENABLED.value)
-
-    sl_pos = arm.read("Present_Position", "shoulder_lift")
-    arm.write("Goal_Position", sl_pos - 1024 - 450, "shoulder_lift")
-    ef_pos = arm.read("Present_Position", "elbow_flex")
-    arm.write("Goal_Position", ef_pos + 1024 + 450, "elbow_flex")
-    time.sleep(2)
-
-    calib = {}
-
-    print("Calibrate shoulder_pan")
-    calib["shoulder_pan"] = move_to_calibrate(arm, "shoulder_pan")
-    arm.write("Goal_Position", calib["shoulder_pan"]["zero_pos"], "shoulder_pan")
-    time.sleep(1)
-
-    print("Calibrate gripper")
-    calib["gripper"] = move_to_calibrate(arm, "gripper", invert_drive_mode=True)
-    time.sleep(1)
-
-    print("Calibrate wrist_flex")
-    calib["wrist_flex"] = move_to_calibrate(arm, "wrist_flex", invert_drive_mode=True)
-    calib["wrist_flex"] = apply_offset(calib["wrist_flex"], offset=-210 + 1024)
-
-    wr_pos = arm.read("Present_Position", "wrist_roll")
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"] - 1024, "wrist_flex")
-    time.sleep(1)
-    arm.write("Goal_Position", wr_pos - 1024, "wrist_roll")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"] - 2048, "wrist_flex")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["gripper"]["end_pos"], "gripper")
-    time.sleep(1)
-
-    print("Calibrate wrist_roll")
-    calib["wrist_roll"] = move_to_calibrate(arm, "wrist_roll", invert_drive_mode=True)
-    calib["wrist_roll"] = apply_offset(calib["wrist_roll"], offset=790)
-
-    arm.write("Goal_Position", calib["wrist_roll"]["zero_pos"] - 1024, "wrist_roll")
-    arm.write("Goal_Position", calib["gripper"]["start_pos"], "gripper")
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"] - 1024, "wrist_flex")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["wrist_roll"]["zero_pos"], "wrist_roll")
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"] - 2048, "wrist_flex")
-
-    def in_between_move_elbow_flex_hook():
-        nonlocal arm, calib
-        arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"], "wrist_flex")
-
-    print("Calibrate elbow_flex")
-    calib["elbow_flex"] = move_to_calibrate(
-        arm,
-        "elbow_flex",
-        invert_drive_mode=True,
-        in_between_move_hook=in_between_move_elbow_flex_hook,
-    )
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"] - 1024, "wrist_flex")
-
-    def in_between_move_shoulder_lift_hook():
-        nonlocal arm, calib
-        sl = arm.read("Present_Position", "shoulder_lift")
-        arm.write("Goal_Position", sl - 1500, "shoulder_lift")
-        time.sleep(1)
-        arm.write("Goal_Position", calib["elbow_flex"]["zero_pos"] + 1536, "elbow_flex")
-        time.sleep(1)
-        arm.write("Goal_Position", calib["wrist_flex"]["start_pos"], "wrist_flex")
-        time.sleep(1)
-
-    print("Calibrate shoulder_lift")
-    calib["shoulder_lift"] = move_to_calibrate(
-        arm, "shoulder_lift", in_between_move_hook=in_between_move_shoulder_lift_hook
-    )
-    calib["shoulder_lift"] = apply_offset(calib["shoulder_lift"], offset=-1024)
-
-    arm.write("Goal_Position", calib["wrist_flex"]["zero_pos"] - 1024, "wrist_flex")
-    time.sleep(1)
-    arm.write("Goal_Position", calib["shoulder_lift"]["zero_pos"] + 2048, "shoulder_lift")
-    arm.write("Goal_Position", calib["elbow_flex"]["zero_pos"] - 1024 - 400, "elbow_flex")
-    time.sleep(2)
-
-    calib_modes = []
-    for name in arm.motor_names:
-        if name == "gripper":
-            calib_modes.append(CalibrationMode.LINEAR.name)
-        else:
-            calib_modes.append(CalibrationMode.DEGREE.name)
-
-    calib_dict = {
-        "homing_offset": [calib[name]["homing_offset"] for name in arm.motor_names],
-        "drive_mode": [calib[name]["drive_mode"] for name in arm.motor_names],
-        "start_pos": [calib[name]["start_pos"] for name in arm.motor_names],
-        "end_pos": [calib[name]["end_pos"] for name in arm.motor_names],
-        "calib_mode": calib_modes,
-        "motor_names": arm.motor_names,
-    }
-
-    # Re-enable original accerlation
-    arm.write("Lock", 0)
-    arm.write("Acceleration", initial_acceleration)
-    time.sleep(1)
-
-    return calib_dict
-
-
-def run_arm_manual_calibration(arm: MotorsBus, robot_type: str, arm_name: str, arm_type: str):
-    """This function ensures that a neural network trained on data collected on a given robot
-    can work on another robot. For instance before calibration, setting a same goal position
-    for each motor of two different robots will get two very different positions. But after calibration,
-    the two robots will move to the same position.To this end, this function computes the homing offset
-    and the drive mode for each motor of a given robot.
-
-    Homing offset is used to shift the motor position to a ]-2048, +2048[ nominal range (when the motor uses 2048 steps
-    to complete a half a turn). This range is set around an arbitrary "zero position" corresponding to all motor positions
-    being 0. During the calibration process, you will need to manually move the robot to this "zero position".
-
-    Drive mode is used to invert the rotation direction of the motor. This is useful when some motors have been assembled
-    in the opposite orientation for some robots. During the calibration process, you will need to manually move the robot
-    to the "rotated position".
-
-    After calibration, the homing offsets and drive modes are stored in a cache.
-
-    Example of usage:
-    ```python
-    run_arm_calibration(arm, "so100", "left", "follower")
-    ```
-    """
-    if (arm.read("Torque_Enable") != TorqueMode.DISABLED.value).any():
-        raise ValueError("To run calibration, the torque must be disabled on all motors.")
-
-    print(f"\nRunning calibration of {robot_type} {arm_name} {arm_type}...")
-
-    reset_middle_positions(arm)
-
-    print("\nMove arm to zero position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="zero"))
-    input("Press Enter to continue...")
-
-    # We arbitrarily chose our zero target position to be a straight horizontal position with gripper upwards and closed.
-    # It is easy to identify and all motors are in a "quarter turn" position. Once calibration is done, this position will
-    # correspond to every motor angle being 0. If you set all 0 as Goal Position, the arm will move in this position.
-    zero_target_pos = convert_degrees_to_steps(ZERO_POSITION_DEGREE, arm.motor_models)
-
-    # Compute homing offset so that `present_position + homing_offset ~= target_position`.
-    zero_pos = arm.read("Present_Position")
-    homing_offset = zero_target_pos - zero_pos
-
-    # The rotated target position corresponds to a rotation of a quarter turn from the zero position.
-    # This allows to identify the rotation direction of each motor.
-    # For instance, if the motor rotates 90 degree, and its value is -90 after applying the homing offset, then we know its rotation direction
-    # is inverted. However, for the calibration being successful, we need everyone to follow the same target position.
-    # Sometimes, there is only one possible rotation direction. For instance, if the gripper is closed, there is only one direction which
-    # corresponds to opening the gripper. When the rotation direction is ambiguous, we arbitrarily rotate clockwise from the point of view
-    # of the previous motor in the kinetic chain.
-    print("\nMove arm to rotated target position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="rotated"))
-    input("Press Enter to continue...")
-
-    rotated_target_pos = convert_degrees_to_steps(ROTATED_POSITION_DEGREE, arm.motor_models)
-
-    # Find drive mode by rotating each motor by a quarter of a turn.
-    # Drive mode indicates if the motor rotation direction should be inverted (=1) or not (=0).
-    rotated_pos = arm.read("Present_Position")
-    drive_mode = (rotated_pos < zero_pos).astype(np.int32)
-
-    # Re-compute homing offset to take into account drive mode
-    rotated_drived_pos = apply_drive_mode(rotated_pos, drive_mode)
-    homing_offset = rotated_target_pos - rotated_drived_pos
-
-    print("\nMove arm to rest position")
-    print("See: " + URL_TEMPLATE.format(robot=robot_type, arm=arm_type, position="rest"))
-    input("Press Enter to continue...")
-    print()
-
-    # Joints with rotational motions are expressed in degrees in nominal range of [-180, 180]
-    calib_modes = []
-    for name in arm.motor_names:
-        if name == "gripper":
-            calib_modes.append(CalibrationMode.LINEAR.name)
-        else:
-            calib_modes.append(CalibrationMode.DEGREE.name)
-
-    calib_dict = {
-        "homing_offset": homing_offset.tolist(),
-        "drive_mode": drive_mode.tolist(),
-        "start_pos": zero_pos.tolist(),
-        "end_pos": rotated_pos.tolist(),
-        "calib_mode": calib_modes,
-        "motor_names": arm.motor_names,
-    }
-    return calib_dict

lerobot/common/robot_devices/robots/lekiwi_remote.py

@@ -1,224 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import base64
-import json
-import threading
-import time
-from pathlib import Path
-
-import cv2
-import zmq
-
-from lerobot.common.robot_devices.robots.mobile_manipulator import LeKiwi
-
-
-def setup_zmq_sockets(config):
-    context = zmq.Context()
-    cmd_socket = context.socket(zmq.PULL)
-    cmd_socket.setsockopt(zmq.CONFLATE, 1)
-    cmd_socket.bind(f"tcp://*:{config.port}")
-
-    video_socket = context.socket(zmq.PUSH)
-    video_socket.setsockopt(zmq.CONFLATE, 1)
-    video_socket.bind(f"tcp://*:{config.video_port}")
-
-    return context, cmd_socket, video_socket
-
-
-def run_camera_capture(cameras, images_lock, latest_images_dict, stop_event):
-    while not stop_event.is_set():
-        local_dict = {}
-        for name, cam in cameras.items():
-            frame = cam.async_read()
-            ret, buffer = cv2.imencode(".jpg", frame, [int(cv2.IMWRITE_JPEG_QUALITY), 90])
-            if ret:
-                local_dict[name] = base64.b64encode(buffer).decode("utf-8")
-            else:
-                local_dict[name] = ""
-        with images_lock:
-            latest_images_dict.update(local_dict)
-        time.sleep(0.01)
-
-
-def calibrate_follower_arm(motors_bus, calib_dir_str):
-    """
-    Calibrates the follower arm. Attempts to load an existing calibration file;
-    if not found, runs manual calibration and saves the result.
-    """
-    calib_dir = Path(calib_dir_str)
-    calib_dir.mkdir(parents=True, exist_ok=True)
-    calib_file = calib_dir / "main_follower.json"
-    try:
-        from lerobot.common.robot_devices.robots.feetech_calibration import run_arm_manual_calibration
-    except ImportError:
-        print("[WARNING] Calibration function not available. Skipping calibration.")
-        return
-
-    if calib_file.exists():
-        with open(calib_file) as f:
-            calibration = json.load(f)
-        print(f"[INFO] Loaded calibration from {calib_file}")
-    else:
-        print("[INFO] Calibration file not found. Running manual calibration...")
-        calibration = run_arm_manual_calibration(motors_bus, "lekiwi", "follower_arm", "follower")
-        print(f"[INFO] Calibration complete. Saving to {calib_file}")
-        with open(calib_file, "w") as f:
-            json.dump(calibration, f)
-    try:
-        motors_bus.set_calibration(calibration)
-        print("[INFO] Applied calibration for follower arm.")
-    except Exception as e:
-        print(f"[WARNING] Could not apply calibration: {e}")
-
-
-def run_lekiwi(robot_config):
-    """
-    Runs the LeKiwi robot:
-      - Sets up cameras and connects them.
-      - Initializes the follower arm motors.
-      - Calibrates the follower arm if necessary.
-      - Creates ZeroMQ sockets for receiving commands and streaming observations.
-      - Processes incoming commands (arm and wheel commands) and sends back sensor and camera data.
-    """
-    # Import helper functions and classes
-    from lerobot.common.robot_devices.cameras.utils import make_cameras_from_configs
-    from lerobot.common.robot_devices.motors.feetech import FeetechMotorsBus, TorqueMode
-
-    # Initialize cameras from the robot configuration.
-    cameras = make_cameras_from_configs(robot_config.cameras)
-    for cam in cameras.values():
-        cam.connect()
-
-    # Initialize the motors bus using the follower arm configuration.
-    motor_config = robot_config.follower_arms.get("main")
-    if motor_config is None:
-        print("[ERROR] Follower arm 'main' configuration not found.")
-        return
-    motors_bus = FeetechMotorsBus(motor_config)
-    motors_bus.connect()
-
-    # Calibrate the follower arm.
-    calibrate_follower_arm(motors_bus, robot_config.calibration_dir)
-
-    # Create the LeKiwi robot instance.
-    robot = LeKiwi(motors_bus)
-
-    # Define the expected arm motor IDs.
-    arm_motor_ids = ["shoulder_pan", "shoulder_lift", "elbow_flex", "wrist_flex", "wrist_roll", "gripper"]
-
-    # Disable torque for each arm motor.
-    for motor in arm_motor_ids:
-        motors_bus.write("Torque_Enable", TorqueMode.DISABLED.value, motor)
-
-    # Set up ZeroMQ sockets.
-    context, cmd_socket, video_socket = setup_zmq_sockets(robot_config)
-
-    # Start the camera capture thread.
-    latest_images_dict = {}
-    images_lock = threading.Lock()
-    stop_event = threading.Event()
-    cam_thread = threading.Thread(
-        target=run_camera_capture, args=(cameras, images_lock, latest_images_dict, stop_event), daemon=True
-    )
-    cam_thread.start()
-
-    last_cmd_time = time.time()
-    print("LeKiwi robot server started. Waiting for commands...")
-
-    try:
-        while True:
-            loop_start_time = time.time()
-
-            # Process incoming commands (non-blocking).
-            while True:
-                try:
-                    msg = cmd_socket.recv_string(zmq.NOBLOCK)
-                except zmq.Again:
-                    break
-                try:
-                    data = json.loads(msg)
-                    # Process arm position commands.
-                    if "arm_positions" in data:
-                        arm_positions = data["arm_positions"]
-                        if not isinstance(arm_positions, list):
-                            print(f"[ERROR] Invalid arm_positions: {arm_positions}")
-                        elif len(arm_positions) < len(arm_motor_ids):
-                            print(
-                                f"[WARNING] Received {len(arm_positions)} arm positions, expected {len(arm_motor_ids)}"
-                            )
-                        else:
-                            for motor, pos in zip(arm_motor_ids, arm_positions, strict=False):
-                                motors_bus.write("Goal_Position", pos, motor)
-                    # Process wheel (base) commands.
-                    if "raw_velocity" in data:
-                        raw_command = data["raw_velocity"]
-                        # Expect keys: "left_wheel", "back_wheel", "right_wheel".
-                        command_speeds = [
-                            int(raw_command.get("left_wheel", 0)),
-                            int(raw_command.get("back_wheel", 0)),
-                            int(raw_command.get("right_wheel", 0)),
-                        ]
-                        robot.set_velocity(command_speeds)
-                        last_cmd_time = time.time()
-                except Exception as e:
-                    print(f"[ERROR] Parsing message failed: {e}")
-
-            # Watchdog: stop the robot if no command is received for over 0.5 seconds.
-            now = time.time()
-            if now - last_cmd_time > 0.5:
-                robot.stop()
-                last_cmd_time = now
-
-            # Read current wheel speeds from the robot.
-            current_velocity = robot.read_velocity()
-
-            # Read the follower arm state from the motors bus.
-            follower_arm_state = []
-            for motor in arm_motor_ids:
-                try:
-                    pos = motors_bus.read("Present_Position", motor)
-                    # Convert the position to a float (or use as is if already numeric).
-                    follower_arm_state.append(float(pos) if not isinstance(pos, (int, float)) else pos)
-                except Exception as e:
-                    print(f"[ERROR] Reading motor {motor} failed: {e}")
-
-            # Get the latest camera images.
-            with images_lock:
-                images_dict_copy = dict(latest_images_dict)
-
-            # Build the observation dictionary.
-            observation = {
-                "images": images_dict_copy,
-                "present_speed": current_velocity,
-                "follower_arm_state": follower_arm_state,
-            }
-            # Send the observation over the video socket.
-            video_socket.send_string(json.dumps(observation))
-
-            # Ensure a short sleep to avoid overloading the CPU.
-            elapsed = time.time() - loop_start_time
-            time.sleep(
-                max(0.033 - elapsed, 0)
-            )  # If robot jitters increase the sleep and monitor cpu load with `top` in cmd
-    except KeyboardInterrupt:
-        print("Shutting down LeKiwi server.")
-    finally:
-        stop_event.set()
-        cam_thread.join()
-        robot.stop()
-        motors_bus.disconnect()
-        cmd_socket.close()
-        video_socket.close()
-        context.term()

lerobot/common/robot_devices/robots/manipulator.py

@@ -1,627 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-"""Contains logic to instantiate a robot, read information from its motors and cameras,
-and send orders to its motors.
-"""
-# TODO(rcadene, aliberts): reorganize the codebase into one file per robot, with the associated
-# calibration procedure, to make it easy for people to add their own robot.
-
-import json
-import logging
-import time
-import warnings
-from pathlib import Path
-
-import numpy as np
-import torch
-
-from lerobot.common.robot_devices.cameras.utils import make_cameras_from_configs
-from lerobot.common.robot_devices.motors.utils import MotorsBus, make_motors_buses_from_configs
-from lerobot.common.robot_devices.robots.configs import ManipulatorRobotConfig
-from lerobot.common.robot_devices.robots.utils import get_arm_id
-from lerobot.common.robot_devices.utils import RobotDeviceAlreadyConnectedError, RobotDeviceNotConnectedError
-
-
-def ensure_safe_goal_position(
-    goal_pos: torch.Tensor, present_pos: torch.Tensor, max_relative_target: float | list[float]
-):
-    # Cap relative action target magnitude for safety.
-    diff = goal_pos - present_pos
-    max_relative_target = torch.tensor(max_relative_target)
-    safe_diff = torch.minimum(diff, max_relative_target)
-    safe_diff = torch.maximum(safe_diff, -max_relative_target)
-    safe_goal_pos = present_pos + safe_diff
-
-    if not torch.allclose(goal_pos, safe_goal_pos):
-        logging.warning(
-            "Relative goal position magnitude had to be clamped to be safe.\n"
-            f"  requested relative goal position target: {diff}\n"
-            f"    clamped relative goal position target: {safe_diff}"
-        )
-
-    return safe_goal_pos
-
-
-class ManipulatorRobot:
-    # TODO(rcadene): Implement force feedback
-    """This class allows to control any manipulator robot of various number of motors.
-
-    Non exhaustive list of robots:
-    - [Koch v1.0](https://github.com/AlexanderKoch-Koch/low_cost_robot), with and without the wrist-to-elbow expansion, developed
-    by Alexander Koch from [Tau Robotics](https://tau-robotics.com)
-    - [Koch v1.1](https://github.com/jess-moss/koch-v1-1) developed by Jess Moss
-    - [Aloha](https://www.trossenrobotics.com/aloha-kits) developed by Trossen Robotics
-
-    Example of instantiation, a pre-defined robot config is required:
-    ```python
-    robot = ManipulatorRobot(KochRobotConfig())
-    ```
-
-    Example of overwriting motors during instantiation:
-    ```python
-    # Defines how to communicate with the motors of the leader and follower arms
-    leader_arms = {
-        "main": DynamixelMotorsBusConfig(
-            port="/dev/tty.usbmodem575E0031751",
-            motors={
-                # name: (index, model)
-                "shoulder_pan": (1, "xl330-m077"),
-                "shoulder_lift": (2, "xl330-m077"),
-                "elbow_flex": (3, "xl330-m077"),
-                "wrist_flex": (4, "xl330-m077"),
-                "wrist_roll": (5, "xl330-m077"),
-                "gripper": (6, "xl330-m077"),
-            },
-        ),
-    }
-    follower_arms = {
-        "main": DynamixelMotorsBusConfig(
-            port="/dev/tty.usbmodem575E0032081",
-            motors={
-                # name: (index, model)
-                "shoulder_pan": (1, "xl430-w250"),
-                "shoulder_lift": (2, "xl430-w250"),
-                "elbow_flex": (3, "xl330-m288"),
-                "wrist_flex": (4, "xl330-m288"),
-                "wrist_roll": (5, "xl330-m288"),
-                "gripper": (6, "xl330-m288"),
-            },
-        ),
-    }
-    robot_config = KochRobotConfig(leader_arms=leader_arms, follower_arms=follower_arms)
-    robot = ManipulatorRobot(robot_config)
-    ```
-
-    Example of overwriting cameras during instantiation:
-    ```python
-    # Defines how to communicate with 2 cameras connected to the computer.
-    # Here, the webcam of the laptop and the phone (connected in USB to the laptop)
-    # can be reached respectively using the camera indices 0 and 1. These indices can be
-    # arbitrary. See the documentation of `OpenCVCamera` to find your own camera indices.
-    cameras = {
-        "laptop": OpenCVCamera(camera_index=0, fps=30, width=640, height=480),
-        "phone": OpenCVCamera(camera_index=1, fps=30, width=640, height=480),
-    }
-    robot = ManipulatorRobot(KochRobotConfig(cameras=cameras))
-    ```
-
-    Once the robot is instantiated, connect motors buses and cameras if any (Required):
-    ```python
-    robot.connect()
-    ```
-
-    Example of highest frequency teleoperation, which doesn't require cameras:
-    ```python
-    while True:
-        robot.teleop_step()
-    ```
-
-    Example of highest frequency data collection from motors and cameras (if any):
-    ```python
-    while True:
-        observation, action = robot.teleop_step(record_data=True)
-    ```
-
-    Example of controlling the robot with a policy:
-    ```python
-    while True:
-        # Uses the follower arms and cameras to capture an observation
-        observation = robot.capture_observation()
-
-        # Assumes a policy has been instantiated
-        with torch.inference_mode():
-            action = policy.select_action(observation)
-
-        # Orders the robot to move
-        robot.send_action(action)
-    ```
-
-    Example of disconnecting which is not mandatory since we disconnect when the object is deleted:
-    ```python
-    robot.disconnect()
-    ```
-    """
-
-    def __init__(
-        self,
-        config: ManipulatorRobotConfig,
-    ):
-        self.config = config
-        self.robot_type = self.config.type
-        self.calibration_dir = Path(self.config.calibration_dir)
-        self.leader_arms = make_motors_buses_from_configs(self.config.leader_arms)
-        self.follower_arms = make_motors_buses_from_configs(self.config.follower_arms)
-        self.cameras = make_cameras_from_configs(self.config.cameras)
-        self.is_connected = False
-        self.logs = {}
-
-    def get_motor_names(self, arm: dict[str, MotorsBus]) -> list:
-        return [f"{arm}_{motor}" for arm, bus in arm.items() for motor in bus.motors]
-
-    @property
-    def camera_features(self) -> dict:
-        cam_ft = {}
-        for cam_key, cam in self.cameras.items():
-            key = f"observation.images.{cam_key}"
-            cam_ft[key] = {
-                "shape": (cam.height, cam.width, cam.channels),
-                "names": ["height", "width", "channels"],
-                "info": None,
-            }
-        return cam_ft
-
-    @property
-    def motor_features(self) -> dict:
-        action_names = self.get_motor_names(self.leader_arms)
-        state_names = self.get_motor_names(self.leader_arms)
-        return {
-            "action": {
-                "dtype": "float32",
-                "shape": (len(action_names),),
-                "names": action_names,
-            },
-            "observation.state": {
-                "dtype": "float32",
-                "shape": (len(state_names),),
-                "names": state_names,
-            },
-        }
-
-    @property
-    def features(self):
-        return {**self.motor_features, **self.camera_features}
-
-    @property
-    def has_camera(self):
-        return len(self.cameras) > 0
-
-    @property
-    def num_cameras(self):
-        return len(self.cameras)
-
-    @property
-    def available_arms(self):
-        available_arms = []
-        for name in self.follower_arms:
-            arm_id = get_arm_id(name, "follower")
-            available_arms.append(arm_id)
-        for name in self.leader_arms:
-            arm_id = get_arm_id(name, "leader")
-            available_arms.append(arm_id)
-        return available_arms
-
-    def connect(self):
-        if self.is_connected:
-            raise RobotDeviceAlreadyConnectedError(
-                "ManipulatorRobot is already connected. Do not run `robot.connect()` twice."
-            )
-
-        if not self.leader_arms and not self.follower_arms and not self.cameras:
-            raise ValueError(
-                "ManipulatorRobot doesn't have any device to connect. See example of usage in docstring of the class."
-            )
-
-        # Connect the arms
-        for name in self.follower_arms:
-            print(f"Connecting {name} follower arm.")
-            self.follower_arms[name].connect()
-        for name in self.leader_arms:
-            print(f"Connecting {name} leader arm.")
-            self.leader_arms[name].connect()
-
-        if self.robot_type in ["koch", "koch_bimanual", "aloha"]:
-            from lerobot.common.robot_devices.motors.dynamixel import TorqueMode
-        elif self.robot_type in ["so100", "so101", "moss", "lekiwi"]:
-            from lerobot.common.robot_devices.motors.feetech import TorqueMode
-
-        # We assume that at connection time, arms are in a rest position, and torque can
-        # be safely disabled to run calibration and/or set robot preset configurations.
-        for name in self.follower_arms:
-            self.follower_arms[name].write("Torque_Enable", TorqueMode.DISABLED.value)
-        for name in self.leader_arms:
-            self.leader_arms[name].write("Torque_Enable", TorqueMode.DISABLED.value)
-
-        self.activate_calibration()
-
-        # Set robot preset (e.g. torque in leader gripper for Koch v1.1)
-        if self.robot_type in ["koch", "koch_bimanual"]:
-            self.set_koch_robot_preset()
-        elif self.robot_type == "aloha":
-            self.set_aloha_robot_preset()
-        elif self.robot_type in ["so100", "so101", "moss", "lekiwi"]:
-            self.set_so100_robot_preset()
-
-        # Enable torque on all motors of the follower arms
-        for name in self.follower_arms:
-            print(f"Activating torque on {name} follower arm.")
-            self.follower_arms[name].write("Torque_Enable", 1)
-
-        if self.config.gripper_open_degree is not None:
-            if self.robot_type not in ["koch", "koch_bimanual"]:
-                raise NotImplementedError(
-                    f"{self.robot_type} does not support position AND current control in the handle, which is require to set the gripper open."
-                )
-            # Set the leader arm in torque mode with the gripper motor set to an angle. This makes it possible
-            # to squeeze the gripper and have it spring back to an open position on its own.
-            for name in self.leader_arms:
-                self.leader_arms[name].write("Torque_Enable", 1, "gripper")
-                self.leader_arms[name].write("Goal_Position", self.config.gripper_open_degree, "gripper")
-
-        # Check both arms can be read
-        for name in self.follower_arms:
-            self.follower_arms[name].read("Present_Position")
-        for name in self.leader_arms:
-            self.leader_arms[name].read("Present_Position")
-
-        # Connect the cameras
-        for name in self.cameras:
-            self.cameras[name].connect()
-
-        self.is_connected = True
-
-    def activate_calibration(self):
-        """After calibration all motors function in human interpretable ranges.
-        Rotations are expressed in degrees in nominal range of [-180, 180],
-        and linear motions (like gripper of Aloha) in nominal range of [0, 100].
-        """
-
-        def load_or_run_calibration_(name, arm, arm_type):
-            arm_id = get_arm_id(name, arm_type)
-            arm_calib_path = self.calibration_dir / f"{arm_id}.json"
-
-            if arm_calib_path.exists():
-                with open(arm_calib_path) as f:
-                    calibration = json.load(f)
-            else:
-                # TODO(rcadene): display a warning in __init__ if calibration file not available
-                print(f"Missing calibration file '{arm_calib_path}'")
-
-                if self.robot_type in ["koch", "koch_bimanual", "aloha"]:
-                    from lerobot.common.robot_devices.robots.dynamixel_calibration import run_arm_calibration
-
-                    calibration = run_arm_calibration(arm, self.robot_type, name, arm_type)
-
-                elif self.robot_type in ["so100", "so101", "moss", "lekiwi"]:
-                    from lerobot.common.robot_devices.robots.feetech_calibration import (
-                        run_arm_manual_calibration,
-                    )
-
-                    calibration = run_arm_manual_calibration(arm, self.robot_type, name, arm_type)
-
-                print(f"Calibration is done! Saving calibration file '{arm_calib_path}'")
-                arm_calib_path.parent.mkdir(parents=True, exist_ok=True)
-                with open(arm_calib_path, "w") as f:
-                    json.dump(calibration, f)
-
-            return calibration
-
-        for name, arm in self.follower_arms.items():
-            calibration = load_or_run_calibration_(name, arm, "follower")
-            arm.set_calibration(calibration)
-        for name, arm in self.leader_arms.items():
-            calibration = load_or_run_calibration_(name, arm, "leader")
-            arm.set_calibration(calibration)
-
-    def set_koch_robot_preset(self):
-        def set_operating_mode_(arm):
-            from lerobot.common.robot_devices.motors.dynamixel import TorqueMode
-
-            if (arm.read("Torque_Enable") != TorqueMode.DISABLED.value).any():
-                raise ValueError("To run set robot preset, the torque must be disabled on all motors.")
-
-            # Use 'extended position mode' for all motors except gripper, because in joint mode the servos can't
-            # rotate more than 360 degrees (from 0 to 4095) And some mistake can happen while assembling the arm,
-            # you could end up with a servo with a position 0 or 4095 at a crucial point See [
-            # https://emanual.robotis.com/docs/en/dxl/x/x_series/#operating-mode11]
-            all_motors_except_gripper = [name for name in arm.motor_names if name != "gripper"]
-            if len(all_motors_except_gripper) > 0:
-                # 4 corresponds to Extended Position on Koch motors
-                arm.write("Operating_Mode", 4, all_motors_except_gripper)
-
-            # Use 'position control current based' for gripper to be limited by the limit of the current.
-            # For the follower gripper, it means it can grasp an object without forcing too much even tho,
-            # it's goal position is a complete grasp (both gripper fingers are ordered to join and reach a touch).
-            # For the leader gripper, it means we can use it as a physical trigger, since we can force with our finger
-            # to make it move, and it will move back to its original target position when we release the force.
-            # 5 corresponds to Current Controlled Position on Koch gripper motors "xl330-m077, xl330-m288"
-            arm.write("Operating_Mode", 5, "gripper")
-
-        for name in self.follower_arms:
-            set_operating_mode_(self.follower_arms[name])
-
-            # Set better PID values to close the gap between recorded states and actions
-            # TODO(rcadene): Implement an automatic procedure to set optimal PID values for each motor
-            self.follower_arms[name].write("Position_P_Gain", 1500, "elbow_flex")
-            self.follower_arms[name].write("Position_I_Gain", 0, "elbow_flex")
-            self.follower_arms[name].write("Position_D_Gain", 600, "elbow_flex")
-
-        if self.config.gripper_open_degree is not None:
-            for name in self.leader_arms:
-                set_operating_mode_(self.leader_arms[name])
-
-                # Enable torque on the gripper of the leader arms, and move it to 45 degrees,
-                # so that we can use it as a trigger to close the gripper of the follower arms.
-                self.leader_arms[name].write("Torque_Enable", 1, "gripper")
-                self.leader_arms[name].write("Goal_Position", self.config.gripper_open_degree, "gripper")
-
-    def set_aloha_robot_preset(self):
-        def set_shadow_(arm):
-            # Set secondary/shadow ID for shoulder and elbow. These joints have two motors.
-            # As a result, if only one of them is required to move to a certain position,
-            # the other will follow. This is to avoid breaking the motors.
-            if "shoulder_shadow" in arm.motor_names:
-                shoulder_idx = arm.read("ID", "shoulder")
-                arm.write("Secondary_ID", shoulder_idx, "shoulder_shadow")
-
-            if "elbow_shadow" in arm.motor_names:
-                elbow_idx = arm.read("ID", "elbow")
-                arm.write("Secondary_ID", elbow_idx, "elbow_shadow")
-
-        for name in self.follower_arms:
-            set_shadow_(self.follower_arms[name])
-
-        for name in self.leader_arms:
-            set_shadow_(self.leader_arms[name])
-
-        for name in self.follower_arms:
-            # Set a velocity limit of 131 as advised by Trossen Robotics
-            self.follower_arms[name].write("Velocity_Limit", 131)
-
-            # Use 'extended position mode' for all motors except gripper, because in joint mode the servos can't
-            # rotate more than 360 degrees (from 0 to 4095) And some mistake can happen while assembling the arm,
-            # you could end up with a servo with a position 0 or 4095 at a crucial point See [
-            # https://emanual.robotis.com/docs/en/dxl/x/x_series/#operating-mode11]
-            all_motors_except_gripper = [
-                name for name in self.follower_arms[name].motor_names if name != "gripper"
-            ]
-            if len(all_motors_except_gripper) > 0:
-                # 4 corresponds to Extended Position on Aloha motors
-                self.follower_arms[name].write("Operating_Mode", 4, all_motors_except_gripper)
-
-            # Use 'position control current based' for follower gripper to be limited by the limit of the current.
-            # It can grasp an object without forcing too much even tho,
-            # it's goal position is a complete grasp (both gripper fingers are ordered to join and reach a touch).
-            # 5 corresponds to Current Controlled Position on Aloha gripper follower "xm430-w350"
-            self.follower_arms[name].write("Operating_Mode", 5, "gripper")
-
-            # Note: We can't enable torque on the leader gripper since "xc430-w150" doesn't have
-            # a Current Controlled Position mode.
-
-        if self.config.gripper_open_degree is not None:
-            warnings.warn(
-                f"`gripper_open_degree` is set to {self.config.gripper_open_degree}, but None is expected for Aloha instead",
-                stacklevel=1,
-            )
-
-    def set_so100_robot_preset(self):
-        for name in self.follower_arms:
-            # Mode=0 for Position Control
-            self.follower_arms[name].write("Mode", 0)
-            # Set P_Coefficient to lower value to avoid shakiness (Default is 32)
-            self.follower_arms[name].write("P_Coefficient", 16)
-            # Set I_Coefficient and D_Coefficient to default value 0 and 32
-            self.follower_arms[name].write("I_Coefficient", 0)
-            self.follower_arms[name].write("D_Coefficient", 32)
-            # Close the write lock so that Maximum_Acceleration gets written to EPROM address,
-            # which is mandatory for Maximum_Acceleration to take effect after rebooting.
-            self.follower_arms[name].write("Lock", 0)
-            # Set Maximum_Acceleration to 254 to speedup acceleration and deceleration of
-            # the motors. Note: this configuration is not in the official STS3215 Memory Table
-            self.follower_arms[name].write("Maximum_Acceleration", 254)
-            self.follower_arms[name].write("Acceleration", 254)
-
-    def teleop_step(
-        self, record_data=False
-    ) -> None | tuple[dict[str, torch.Tensor], dict[str, torch.Tensor]]:
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                "ManipulatorRobot is not connected. You need to run `robot.connect()`."
-            )
-
-        # Prepare to assign the position of the leader to the follower
-        leader_pos = {}
-        for name in self.leader_arms:
-            before_lread_t = time.perf_counter()
-            leader_pos[name] = self.leader_arms[name].read("Present_Position")
-            leader_pos[name] = torch.from_numpy(leader_pos[name])
-            self.logs[f"read_leader_{name}_pos_dt_s"] = time.perf_counter() - before_lread_t
-
-        # Send goal position to the follower
-        follower_goal_pos = {}
-        for name in self.follower_arms:
-            before_fwrite_t = time.perf_counter()
-            goal_pos = leader_pos[name]
-
-            # Cap goal position when too far away from present position.
-            # Slower fps expected due to reading from the follower.
-            if self.config.max_relative_target is not None:
-                present_pos = self.follower_arms[name].read("Present_Position")
-                present_pos = torch.from_numpy(present_pos)
-                goal_pos = ensure_safe_goal_position(goal_pos, present_pos, self.config.max_relative_target)
-
-            # Used when record_data=True
-            follower_goal_pos[name] = goal_pos
-
-            goal_pos = goal_pos.numpy().astype(np.float32)
-            self.follower_arms[name].write("Goal_Position", goal_pos)
-            self.logs[f"write_follower_{name}_goal_pos_dt_s"] = time.perf_counter() - before_fwrite_t
-
-        # Early exit when recording data is not requested
-        if not record_data:
-            return
-
-        # TODO(rcadene): Add velocity and other info
-        # Read follower position
-        follower_pos = {}
-        for name in self.follower_arms:
-            before_fread_t = time.perf_counter()
-            follower_pos[name] = self.follower_arms[name].read("Present_Position")
-            follower_pos[name] = torch.from_numpy(follower_pos[name])
-            self.logs[f"read_follower_{name}_pos_dt_s"] = time.perf_counter() - before_fread_t
-
-        # Create state by concatenating follower current position
-        state = []
-        for name in self.follower_arms:
-            if name in follower_pos:
-                state.append(follower_pos[name])
-        state = torch.cat(state)
-
-        # Create action by concatenating follower goal position
-        action = []
-        for name in self.follower_arms:
-            if name in follower_goal_pos:
-                action.append(follower_goal_pos[name])
-        action = torch.cat(action)
-
-        # Capture images from cameras
-        images = {}
-        for name in self.cameras:
-            before_camread_t = time.perf_counter()
-            images[name] = self.cameras[name].async_read()
-            images[name] = torch.from_numpy(images[name])
-            self.logs[f"read_camera_{name}_dt_s"] = self.cameras[name].logs["delta_timestamp_s"]
-            self.logs[f"async_read_camera_{name}_dt_s"] = time.perf_counter() - before_camread_t
-
-        # Populate output dictionaries
-        obs_dict, action_dict = {}, {}
-        obs_dict["observation.state"] = state
-        action_dict["action"] = action
-        for name in self.cameras:
-            obs_dict[f"observation.images.{name}"] = images[name]
-
-        return obs_dict, action_dict
-
-    def capture_observation(self):
-        """The returned observations do not have a batch dimension."""
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                "ManipulatorRobot is not connected. You need to run `robot.connect()`."
-            )
-
-        # Read follower position
-        follower_pos = {}
-        for name in self.follower_arms:
-            before_fread_t = time.perf_counter()
-            follower_pos[name] = self.follower_arms[name].read("Present_Position")
-            follower_pos[name] = torch.from_numpy(follower_pos[name])
-            self.logs[f"read_follower_{name}_pos_dt_s"] = time.perf_counter() - before_fread_t
-
-        # Create state by concatenating follower current position
-        state = []
-        for name in self.follower_arms:
-            if name in follower_pos:
-                state.append(follower_pos[name])
-        state = torch.cat(state)
-
-        # Capture images from cameras
-        images = {}
-        for name in self.cameras:
-            before_camread_t = time.perf_counter()
-            images[name] = self.cameras[name].async_read()
-            images[name] = torch.from_numpy(images[name])
-            self.logs[f"read_camera_{name}_dt_s"] = self.cameras[name].logs["delta_timestamp_s"]
-            self.logs[f"async_read_camera_{name}_dt_s"] = time.perf_counter() - before_camread_t
-
-        # Populate output dictionaries and format to pytorch
-        obs_dict = {}
-        obs_dict["observation.state"] = state
-        for name in self.cameras:
-            obs_dict[f"observation.images.{name}"] = images[name]
-        return obs_dict
-
-    def send_action(self, action: torch.Tensor) -> torch.Tensor:
-        """Command the follower arms to move to a target joint configuration.
-
-        The relative action magnitude may be clipped depending on the configuration parameter
-        `max_relative_target`. In this case, the action sent differs from original action.
-        Thus, this function always returns the action actually sent.
-
-        Args:
-            action: tensor containing the concatenated goal positions for the follower arms.
-        """
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                "ManipulatorRobot is not connected. You need to run `robot.connect()`."
-            )
-
-        from_idx = 0
-        to_idx = 0
-        action_sent = []
-        for name in self.follower_arms:
-            # Get goal position of each follower arm by splitting the action vector
-            to_idx += len(self.follower_arms[name].motor_names)
-            goal_pos = action[from_idx:to_idx]
-            from_idx = to_idx
-
-            # Cap goal position when too far away from present position.
-            # Slower fps expected due to reading from the follower.
-            if self.config.max_relative_target is not None:
-                present_pos = self.follower_arms[name].read("Present_Position")
-                present_pos = torch.from_numpy(present_pos)
-                goal_pos = ensure_safe_goal_position(goal_pos, present_pos, self.config.max_relative_target)
-
-            # Save tensor to concat and return
-            action_sent.append(goal_pos)
-
-            # Send goal position to each follower
-            goal_pos = goal_pos.numpy().astype(np.float32)
-            self.follower_arms[name].write("Goal_Position", goal_pos)
-
-        return torch.cat(action_sent)
-
-    def print_logs(self):
-        pass
-        # TODO(aliberts): move robot-specific logs logic here
-
-    def disconnect(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError(
-                "ManipulatorRobot is not connected. You need to run `robot.connect()` before disconnecting."
-            )
-
-        for name in self.follower_arms:
-            self.follower_arms[name].disconnect()
-
-        for name in self.leader_arms:
-            self.leader_arms[name].disconnect()
-
-        for name in self.cameras:
-            self.cameras[name].disconnect()
-
-        self.is_connected = False
-
-    def __del__(self):
-        if getattr(self, "is_connected", False):
-            self.disconnect()

lerobot/common/robot_devices/robots/mobile_manipulator.py

@@ -1,703 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import base64
-import json
-import os
-import sys
-from pathlib import Path
-
-import cv2
-import numpy as np
-import torch
-import zmq
-
-from lerobot.common.robot_devices.cameras.utils import make_cameras_from_configs
-from lerobot.common.robot_devices.motors.feetech import TorqueMode
-from lerobot.common.robot_devices.motors.utils import MotorsBus, make_motors_buses_from_configs
-from lerobot.common.robot_devices.robots.configs import LeKiwiRobotConfig
-from lerobot.common.robot_devices.robots.feetech_calibration import run_arm_manual_calibration
-from lerobot.common.robot_devices.robots.utils import get_arm_id
-from lerobot.common.robot_devices.utils import RobotDeviceNotConnectedError
-
-PYNPUT_AVAILABLE = True
-try:
-    # Only import if there's a valid X server or if we're not on a Pi
-    if ("DISPLAY" not in os.environ) and ("linux" in sys.platform):
-        print("No DISPLAY set. Skipping pynput import.")
-        raise ImportError("pynput blocked intentionally due to no display.")
-
-    from pynput import keyboard
-except ImportError:
-    keyboard = None
-    PYNPUT_AVAILABLE = False
-except Exception as e:
-    keyboard = None
-    PYNPUT_AVAILABLE = False
-    print(f"Could not import pynput: {e}")
-
-
-class MobileManipulator:
-    """
-    MobileManipulator is a class for connecting to and controlling a remote mobile manipulator robot.
-    The robot includes a three omniwheel mobile base and a remote follower arm.
-    The leader arm is connected locally (on the laptop) and its joint positions are recorded and then
-    forwarded to the remote follower arm (after applying a safety clamp).
-    In parallel, keyboard teleoperation is used to generate raw velocity commands for the wheels.
-    """
-
-    def __init__(self, config: LeKiwiRobotConfig):
-        """
-        Expected keys in config:
-          - ip, port, video_port for the remote connection.
-          - calibration_dir, leader_arms, follower_arms, max_relative_target, etc.
-        """
-        self.robot_type = config.type
-        self.config = config
-        self.remote_ip = config.ip
-        self.remote_port = config.port
-        self.remote_port_video = config.video_port
-        self.calibration_dir = Path(self.config.calibration_dir)
-        self.logs = {}
-
-        self.teleop_keys = self.config.teleop_keys
-
-        # For teleoperation, the leader arm (local) is used to record the desired arm pose.
-        self.leader_arms = make_motors_buses_from_configs(self.config.leader_arms)
-
-        self.follower_arms = make_motors_buses_from_configs(self.config.follower_arms)
-
-        self.cameras = make_cameras_from_configs(self.config.cameras)
-
-        self.is_connected = False
-
-        self.last_frames = {}
-        self.last_present_speed = {}
-        self.last_remote_arm_state = torch.zeros(6, dtype=torch.float32)
-
-        # Define three speed levels and a current index
-        self.speed_levels = [
-            {"xy": 0.1, "theta": 30},  # slow
-            {"xy": 0.2, "theta": 60},  # medium
-            {"xy": 0.3, "theta": 90},  # fast
-        ]
-        self.speed_index = 0  # Start at slow
-
-        # ZeroMQ context and sockets.
-        self.context = None
-        self.cmd_socket = None
-        self.video_socket = None
-
-        # Keyboard state for base teleoperation.
-        self.running = True
-        self.pressed_keys = {
-            "forward": False,
-            "backward": False,
-            "left": False,
-            "right": False,
-            "rotate_left": False,
-            "rotate_right": False,
-        }
-
-        if PYNPUT_AVAILABLE:
-            print("pynput is available - enabling local keyboard listener.")
-            self.listener = keyboard.Listener(
-                on_press=self.on_press,
-                on_release=self.on_release,
-            )
-            self.listener.start()
-        else:
-            print("pynput not available - skipping local keyboard listener.")
-            self.listener = None
-
-    def get_motor_names(self, arms: dict[str, MotorsBus]) -> list:
-        return [f"{arm}_{motor}" for arm, bus in arms.items() for motor in bus.motors]
-
-    @property
-    def camera_features(self) -> dict:
-        cam_ft = {}
-        for cam_key, cam in self.cameras.items():
-            key = f"observation.images.{cam_key}"
-            cam_ft[key] = {
-                "shape": (cam.height, cam.width, cam.channels),
-                "names": ["height", "width", "channels"],
-                "info": None,
-            }
-        return cam_ft
-
-    @property
-    def motor_features(self) -> dict:
-        follower_arm_names = [
-            "shoulder_pan",
-            "shoulder_lift",
-            "elbow_flex",
-            "wrist_flex",
-            "wrist_roll",
-            "gripper",
-        ]
-        observations = ["x_mm", "y_mm", "theta"]
-        combined_names = follower_arm_names + observations
-        return {
-            "action": {
-                "dtype": "float32",
-                "shape": (len(combined_names),),
-                "names": combined_names,
-            },
-            "observation.state": {
-                "dtype": "float32",
-                "shape": (len(combined_names),),
-                "names": combined_names,
-            },
-        }
-
-    @property
-    def features(self):
-        return {**self.motor_features, **self.camera_features}
-
-    @property
-    def has_camera(self):
-        return len(self.cameras) > 0
-
-    @property
-    def num_cameras(self):
-        return len(self.cameras)
-
-    @property
-    def available_arms(self):
-        available = []
-        for name in self.leader_arms:
-            available.append(get_arm_id(name, "leader"))
-        for name in self.follower_arms:
-            available.append(get_arm_id(name, "follower"))
-        return available
-
-    def on_press(self, key):
-        try:
-            # Movement
-            if key.char == self.teleop_keys["forward"]:
-                self.pressed_keys["forward"] = True
-            elif key.char == self.teleop_keys["backward"]:
-                self.pressed_keys["backward"] = True
-            elif key.char == self.teleop_keys["left"]:
-                self.pressed_keys["left"] = True
-            elif key.char == self.teleop_keys["right"]:
-                self.pressed_keys["right"] = True
-            elif key.char == self.teleop_keys["rotate_left"]:
-                self.pressed_keys["rotate_left"] = True
-            elif key.char == self.teleop_keys["rotate_right"]:
-                self.pressed_keys["rotate_right"] = True
-
-            # Quit teleoperation
-            elif key.char == self.teleop_keys["quit"]:
-                self.running = False
-                return False
-
-            # Speed control
-            elif key.char == self.teleop_keys["speed_up"]:
-                self.speed_index = min(self.speed_index + 1, 2)
-                print(f"Speed index increased to {self.speed_index}")
-            elif key.char == self.teleop_keys["speed_down"]:
-                self.speed_index = max(self.speed_index - 1, 0)
-                print(f"Speed index decreased to {self.speed_index}")
-
-        except AttributeError:
-            # e.g., if key is special like Key.esc
-            if key == keyboard.Key.esc:
-                self.running = False
-                return False
-
-    def on_release(self, key):
-        try:
-            if hasattr(key, "char"):
-                if key.char == self.teleop_keys["forward"]:
-                    self.pressed_keys["forward"] = False
-                elif key.char == self.teleop_keys["backward"]:
-                    self.pressed_keys["backward"] = False
-                elif key.char == self.teleop_keys["left"]:
-                    self.pressed_keys["left"] = False
-                elif key.char == self.teleop_keys["right"]:
-                    self.pressed_keys["right"] = False
-                elif key.char == self.teleop_keys["rotate_left"]:
-                    self.pressed_keys["rotate_left"] = False
-                elif key.char == self.teleop_keys["rotate_right"]:
-                    self.pressed_keys["rotate_right"] = False
-        except AttributeError:
-            pass
-
-    def connect(self):
-        if not self.leader_arms:
-            raise ValueError("MobileManipulator has no leader arm to connect.")
-        for name in self.leader_arms:
-            print(f"Connecting {name} leader arm.")
-            self.calibrate_leader()
-
-        # Set up ZeroMQ sockets to communicate with the remote mobile robot.
-        self.context = zmq.Context()
-        self.cmd_socket = self.context.socket(zmq.PUSH)
-        connection_string = f"tcp://{self.remote_ip}:{self.remote_port}"
-        self.cmd_socket.connect(connection_string)
-        self.cmd_socket.setsockopt(zmq.CONFLATE, 1)
-        self.video_socket = self.context.socket(zmq.PULL)
-        video_connection = f"tcp://{self.remote_ip}:{self.remote_port_video}"
-        self.video_socket.connect(video_connection)
-        self.video_socket.setsockopt(zmq.CONFLATE, 1)
-        print(
-            f"[INFO] Connected to remote robot at {connection_string} and video stream at {video_connection}."
-        )
-        self.is_connected = True
-
-    def load_or_run_calibration_(self, name, arm, arm_type):
-        arm_id = get_arm_id(name, arm_type)
-        arm_calib_path = self.calibration_dir / f"{arm_id}.json"
-
-        if arm_calib_path.exists():
-            with open(arm_calib_path) as f:
-                calibration = json.load(f)
-        else:
-            print(f"Missing calibration file '{arm_calib_path}'")
-            calibration = run_arm_manual_calibration(arm, self.robot_type, name, arm_type)
-            print(f"Calibration is done! Saving calibration file '{arm_calib_path}'")
-            arm_calib_path.parent.mkdir(parents=True, exist_ok=True)
-            with open(arm_calib_path, "w") as f:
-                json.dump(calibration, f)
-
-        return calibration
-
-    def calibrate_leader(self):
-        for name, arm in self.leader_arms.items():
-            # Connect the bus
-            arm.connect()
-
-            # Disable torque on all motors
-            for motor_id in arm.motors:
-                arm.write("Torque_Enable", TorqueMode.DISABLED.value, motor_id)
-
-            # Now run calibration
-            calibration = self.load_or_run_calibration_(name, arm, "leader")
-            arm.set_calibration(calibration)
-
-    def calibrate_follower(self):
-        for name, bus in self.follower_arms.items():
-            bus.connect()
-
-            # Disable torque on all motors
-            for motor_id in bus.motors:
-                bus.write("Torque_Enable", 0, motor_id)
-
-            # Then filter out wheels
-            arm_only_dict = {k: v for k, v in bus.motors.items() if not k.startswith("wheel_")}
-            if not arm_only_dict:
-                continue
-
-            original_motors = bus.motors
-            bus.motors = arm_only_dict
-
-            calibration = self.load_or_run_calibration_(name, bus, "follower")
-            bus.set_calibration(calibration)
-
-            bus.motors = original_motors
-
-    def _get_data(self):
-        """
-        Polls the video socket for up to 15 ms. If data arrives, decode only
-        the *latest* message, returning frames, speed, and arm state. If
-        nothing arrives for any field, use the last known values.
-        """
-        frames = {}
-        present_speed = {}
-        remote_arm_state_tensor = torch.zeros(6, dtype=torch.float32)
-
-        # Poll up to 15 ms
-        poller = zmq.Poller()
-        poller.register(self.video_socket, zmq.POLLIN)
-        socks = dict(poller.poll(15))
-        if self.video_socket not in socks or socks[self.video_socket] != zmq.POLLIN:
-            # No new data arrived → reuse ALL old data
-            return (self.last_frames, self.last_present_speed, self.last_remote_arm_state)
-
-        # Drain all messages, keep only the last
-        last_msg = None
-        while True:
-            try:
-                obs_string = self.video_socket.recv_string(zmq.NOBLOCK)
-                last_msg = obs_string
-            except zmq.Again:
-                break
-
-        if not last_msg:
-            # No new message → also reuse old
-            return (self.last_frames, self.last_present_speed, self.last_remote_arm_state)
-
-        # Decode only the final message
-        try:
-            observation = json.loads(last_msg)
-
-            images_dict = observation.get("images", {})
-            new_speed = observation.get("present_speed", {})
-            new_arm_state = observation.get("follower_arm_state", None)
-
-            # Convert images
-            for cam_name, image_b64 in images_dict.items():
-                if image_b64:
-                    jpg_data = base64.b64decode(image_b64)
-                    np_arr = np.frombuffer(jpg_data, dtype=np.uint8)
-                    frame_candidate = cv2.imdecode(np_arr, cv2.IMREAD_COLOR)
-                    if frame_candidate is not None:
-                        frames[cam_name] = frame_candidate
-
-            # If remote_arm_state is None and frames is None there is no message then use the previous message
-            if new_arm_state is not None and frames is not None:
-                self.last_frames = frames
-
-                remote_arm_state_tensor = torch.tensor(new_arm_state, dtype=torch.float32)
-                self.last_remote_arm_state = remote_arm_state_tensor
-
-                present_speed = new_speed
-                self.last_present_speed = new_speed
-            else:
-                frames = self.last_frames
-
-                remote_arm_state_tensor = self.last_remote_arm_state
-
-                present_speed = self.last_present_speed
-
-        except Exception as e:
-            print(f"[DEBUG] Error decoding video message: {e}")
-            # If decode fails, fall back to old data
-            return (self.last_frames, self.last_present_speed, self.last_remote_arm_state)
-
-        return frames, present_speed, remote_arm_state_tensor
-
-    def _process_present_speed(self, present_speed: dict) -> torch.Tensor:
-        state_tensor = torch.zeros(3, dtype=torch.int32)
-        if present_speed:
-            decoded = {key: MobileManipulator.raw_to_degps(value) for key, value in present_speed.items()}
-            if "1" in decoded:
-                state_tensor[0] = decoded["1"]
-            if "2" in decoded:
-                state_tensor[1] = decoded["2"]
-            if "3" in decoded:
-                state_tensor[2] = decoded["3"]
-        return state_tensor
-
-    def teleop_step(
-        self, record_data: bool = False
-    ) -> None | tuple[dict[str, torch.Tensor], dict[str, torch.Tensor]]:
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError("MobileManipulator is not connected. Run `connect()` first.")
-
-        speed_setting = self.speed_levels[self.speed_index]
-        xy_speed = speed_setting["xy"]  # e.g. 0.1, 0.25, or 0.4
-        theta_speed = speed_setting["theta"]  # e.g. 30, 60, or 90
-
-        # Prepare to assign the position of the leader to the follower
-        arm_positions = []
-        for name in self.leader_arms:
-            pos = self.leader_arms[name].read("Present_Position")
-            pos_tensor = torch.from_numpy(pos).float()
-            arm_positions.extend(pos_tensor.tolist())
-
-        y_cmd = 0.0  # m/s forward/backward
-        x_cmd = 0.0  # m/s lateral
-        theta_cmd = 0.0  # deg/s rotation
-        if self.pressed_keys["forward"]:
-            y_cmd += xy_speed
-        if self.pressed_keys["backward"]:
-            y_cmd -= xy_speed
-        if self.pressed_keys["left"]:
-            x_cmd += xy_speed
-        if self.pressed_keys["right"]:
-            x_cmd -= xy_speed
-        if self.pressed_keys["rotate_left"]:
-            theta_cmd += theta_speed
-        if self.pressed_keys["rotate_right"]:
-            theta_cmd -= theta_speed
-
-        wheel_commands = self.body_to_wheel_raw(x_cmd, y_cmd, theta_cmd)
-
-        message = {"raw_velocity": wheel_commands, "arm_positions": arm_positions}
-        self.cmd_socket.send_string(json.dumps(message))
-
-        if not record_data:
-            return
-
-        obs_dict = self.capture_observation()
-
-        arm_state_tensor = torch.tensor(arm_positions, dtype=torch.float32)
-
-        wheel_velocity_tuple = self.wheel_raw_to_body(wheel_commands)
-        wheel_velocity_mm = (
-            wheel_velocity_tuple[0] * 1000.0,
-            wheel_velocity_tuple[1] * 1000.0,
-            wheel_velocity_tuple[2],
-        )
-        wheel_tensor = torch.tensor(wheel_velocity_mm, dtype=torch.float32)
-        action_tensor = torch.cat([arm_state_tensor, wheel_tensor])
-        action_dict = {"action": action_tensor}
-
-        return obs_dict, action_dict
-
-    def capture_observation(self) -> dict:
-        """
-        Capture observations from the remote robot: current follower arm positions,
-        present wheel speeds (converted to body-frame velocities: x, y, theta),
-        and a camera frame.
-        """
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError("Not connected. Run `connect()` first.")
-
-        frames, present_speed, remote_arm_state_tensor = self._get_data()
-
-        body_state = self.wheel_raw_to_body(present_speed)
-
-        body_state_mm = (body_state[0] * 1000.0, body_state[1] * 1000.0, body_state[2])  # Convert x,y to mm/s
-        wheel_state_tensor = torch.tensor(body_state_mm, dtype=torch.float32)
-        combined_state_tensor = torch.cat((remote_arm_state_tensor, wheel_state_tensor), dim=0)
-
-        obs_dict = {"observation.state": combined_state_tensor}
-
-        # Loop over each configured camera
-        for cam_name, cam in self.cameras.items():
-            frame = frames.get(cam_name, None)
-            if frame is None:
-                # Create a black image using the camera's configured width, height, and channels
-                frame = np.zeros((cam.height, cam.width, cam.channels), dtype=np.uint8)
-            obs_dict[f"observation.images.{cam_name}"] = torch.from_numpy(frame)
-
-        return obs_dict
-
-    def send_action(self, action: torch.Tensor) -> torch.Tensor:
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError("Not connected. Run `connect()` first.")
-
-        # Ensure the action tensor has at least 9 elements:
-        #   - First 6: arm positions.
-        #   - Last 3: base commands.
-        if action.numel() < 9:
-            # Pad with zeros if there are not enough elements.
-            padded = torch.zeros(9, dtype=action.dtype)
-            padded[: action.numel()] = action
-            action = padded
-
-        # Extract arm and base actions.
-        arm_actions = action[:6].flatten()
-        base_actions = action[6:].flatten()
-
-        x_cmd_mm = base_actions[0].item()  # mm/s
-        y_cmd_mm = base_actions[1].item()  # mm/s
-        theta_cmd = base_actions[2].item()  # deg/s
-
-        # Convert mm/s to m/s for the kinematics calculations.
-        x_cmd = x_cmd_mm / 1000.0  # m/s
-        y_cmd = y_cmd_mm / 1000.0  # m/s
-
-        # Compute wheel commands from body commands.
-        wheel_commands = self.body_to_wheel_raw(x_cmd, y_cmd, theta_cmd)
-
-        arm_positions_list = arm_actions.tolist()
-
-        message = {"raw_velocity": wheel_commands, "arm_positions": arm_positions_list}
-        self.cmd_socket.send_string(json.dumps(message))
-
-        return action
-
-    def print_logs(self):
-        pass
-
-    def disconnect(self):
-        if not self.is_connected:
-            raise RobotDeviceNotConnectedError("Not connected.")
-        if self.cmd_socket:
-            stop_cmd = {
-                "raw_velocity": {"left_wheel": 0, "back_wheel": 0, "right_wheel": 0},
-                "arm_positions": {},
-            }
-            self.cmd_socket.send_string(json.dumps(stop_cmd))
-            self.cmd_socket.close()
-        if self.video_socket:
-            self.video_socket.close()
-        if self.context:
-            self.context.term()
-        if PYNPUT_AVAILABLE:
-            self.listener.stop()
-        self.is_connected = False
-        print("[INFO] Disconnected from remote robot.")
-
-    def __del__(self):
-        if getattr(self, "is_connected", False):
-            self.disconnect()
-        if PYNPUT_AVAILABLE:
-            self.listener.stop()
-
-    @staticmethod
-    def degps_to_raw(degps: float) -> int:
-        steps_per_deg = 4096.0 / 360.0
-        speed_in_steps = abs(degps) * steps_per_deg
-        speed_int = int(round(speed_in_steps))
-        if speed_int > 0x7FFF:
-            speed_int = 0x7FFF
-        if degps < 0:
-            return speed_int | 0x8000
-        else:
-            return speed_int & 0x7FFF
-
-    @staticmethod
-    def raw_to_degps(raw_speed: int) -> float:
-        steps_per_deg = 4096.0 / 360.0
-        magnitude = raw_speed & 0x7FFF
-        degps = magnitude / steps_per_deg
-        if raw_speed & 0x8000:
-            degps = -degps
-        return degps
-
-    def body_to_wheel_raw(
-        self,
-        x_cmd: float,
-        y_cmd: float,
-        theta_cmd: float,
-        wheel_radius: float = 0.05,
-        base_radius: float = 0.125,
-        max_raw: int = 3000,
-    ) -> dict:
-        """
-        Convert desired body-frame velocities into wheel raw commands.
-
-        Parameters:
-          x_cmd      : Linear velocity in x (m/s).
-          y_cmd      : Linear velocity in y (m/s).
-          theta_cmd  : Rotational velocity (deg/s).
-          wheel_radius: Radius of each wheel (meters).
-          base_radius : Distance from the center of rotation to each wheel (meters).
-          max_raw    : Maximum allowed raw command (ticks) per wheel.
-
-        Returns:
-          A dictionary with wheel raw commands:
-             {"left_wheel": value, "back_wheel": value, "right_wheel": value}.
-
-        Notes:
-          - Internally, the method converts theta_cmd to rad/s for the kinematics.
-          - The raw command is computed from the wheels angular speed in deg/s
-            using degps_to_raw(). If any command exceeds max_raw, all commands
-            are scaled down proportionally.
-        """
-        # Convert rotational velocity from deg/s to rad/s.
-        theta_rad = theta_cmd * (np.pi / 180.0)
-        # Create the body velocity vector [x, y, theta_rad].
-        velocity_vector = np.array([x_cmd, y_cmd, theta_rad])
-
-        # Define the wheel mounting angles (defined from y axis cw)
-        angles = np.radians(np.array([300, 180, 60]))
-        # Build the kinematic matrix: each row maps body velocities to a wheel’s linear speed.
-        # The third column (base_radius) accounts for the effect of rotation.
-        m = np.array([[np.cos(a), np.sin(a), base_radius] for a in angles])
-
-        # Compute each wheel’s linear speed (m/s) and then its angular speed (rad/s).
-        wheel_linear_speeds = m.dot(velocity_vector)
-        wheel_angular_speeds = wheel_linear_speeds / wheel_radius
-
-        # Convert wheel angular speeds from rad/s to deg/s.
-        wheel_degps = wheel_angular_speeds * (180.0 / np.pi)
-
-        # Scaling
-        steps_per_deg = 4096.0 / 360.0
-        raw_floats = [abs(degps) * steps_per_deg for degps in wheel_degps]
-        max_raw_computed = max(raw_floats)
-        if max_raw_computed > max_raw:
-            scale = max_raw / max_raw_computed
-            wheel_degps = wheel_degps * scale
-
-        # Convert each wheel’s angular speed (deg/s) to a raw integer.
-        wheel_raw = [MobileManipulator.degps_to_raw(deg) for deg in wheel_degps]
-
-        return {"left_wheel": wheel_raw[0], "back_wheel": wheel_raw[1], "right_wheel": wheel_raw[2]}
-
-    def wheel_raw_to_body(
-        self, wheel_raw: dict, wheel_radius: float = 0.05, base_radius: float = 0.125
-    ) -> tuple:
-        """
-        Convert wheel raw command feedback back into body-frame velocities.
-
-        Parameters:
-          wheel_raw   : Dictionary with raw wheel commands (keys: "left_wheel", "back_wheel", "right_wheel").
-          wheel_radius: Radius of each wheel (meters).
-          base_radius : Distance from the robot center to each wheel (meters).
-
-        Returns:
-          A tuple (x_cmd, y_cmd, theta_cmd) where:
-             x_cmd      : Linear velocity in x (m/s).
-             y_cmd      : Linear velocity in y (m/s).
-             theta_cmd  : Rotational velocity in deg/s.
-        """
-        # Extract the raw values in order.
-        raw_list = [
-            int(wheel_raw.get("left_wheel", 0)),
-            int(wheel_raw.get("back_wheel", 0)),
-            int(wheel_raw.get("right_wheel", 0)),
-        ]
-
-        # Convert each raw command back to an angular speed in deg/s.
-        wheel_degps = np.array([MobileManipulator.raw_to_degps(r) for r in raw_list])
-        # Convert from deg/s to rad/s.
-        wheel_radps = wheel_degps * (np.pi / 180.0)
-        # Compute each wheel’s linear speed (m/s) from its angular speed.
-        wheel_linear_speeds = wheel_radps * wheel_radius
-
-        # Define the wheel mounting angles (defined from y axis cw)
-        angles = np.radians(np.array([300, 180, 60]))
-        m = np.array([[np.cos(a), np.sin(a), base_radius] for a in angles])
-
-        # Solve the inverse kinematics: body_velocity = M⁻¹ · wheel_linear_speeds.
-        m_inv = np.linalg.inv(m)
-        velocity_vector = m_inv.dot(wheel_linear_speeds)
-        x_cmd, y_cmd, theta_rad = velocity_vector
-        theta_cmd = theta_rad * (180.0 / np.pi)
-        return (x_cmd, y_cmd, theta_cmd)
-
-
-class LeKiwi:
-    def __init__(self, motor_bus):
-        """
-        Initializes the LeKiwi with Feetech motors bus.
-        """
-        self.motor_bus = motor_bus
-        self.motor_ids = ["left_wheel", "back_wheel", "right_wheel"]
-
-        # Initialize motors in velocity mode.
-        self.motor_bus.write("Lock", 0)
-        self.motor_bus.write("Mode", [1, 1, 1], self.motor_ids)
-        self.motor_bus.write("Lock", 1)
-        print("Motors set to velocity mode.")
-
-    def read_velocity(self):
-        """
-        Reads the raw speeds for all wheels. Returns a dictionary with motor names:
-        """
-        raw_speeds = self.motor_bus.read("Present_Speed", self.motor_ids)
-        return {
-            "left_wheel": int(raw_speeds[0]),
-            "back_wheel": int(raw_speeds[1]),
-            "right_wheel": int(raw_speeds[2]),
-        }
-
-    def set_velocity(self, command_speeds):
-        """
-        Sends raw velocity commands (16-bit encoded values) directly to the motor bus.
-        The order of speeds must correspond to self.motor_ids.
-        """
-        self.motor_bus.write("Goal_Speed", command_speeds, self.motor_ids)
-
-    def stop(self):
-        """Stops the robot by setting all motor speeds to zero."""
-        self.motor_bus.write("Goal_Speed", [0, 0, 0], self.motor_ids)
-        print("Motors stopped.")

lerobot/common/robot_devices/robots/stretch.py

@@ -1,208 +0,0 @@
-#!/usr/bin/env python
-
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import time
-from dataclasses import replace
-
-import torch
-from stretch_body.gamepad_teleop import GamePadTeleop
-from stretch_body.robot import Robot as StretchAPI
-from stretch_body.robot_params import RobotParams
-
-from lerobot.common.robot_devices.robots.configs import StretchRobotConfig
-
-
-class StretchRobot(StretchAPI):
-    """Wrapper of stretch_body.robot.Robot"""
-
-    def __init__(self, config: StretchRobotConfig | None = None, **kwargs):
-        super().__init__()
-        if config is None:
-            self.config = StretchRobotConfig(**kwargs)
-        else:
-            # Overwrite config arguments using kwargs
-            self.config = replace(config, **kwargs)
-
-        self.robot_type = self.config.type
-        self.cameras = self.config.cameras
-        self.is_connected = False
-        self.teleop = None
-        self.logs = {}
-
-        # TODO(aliberts): test this
-        RobotParams.set_logging_level("WARNING")
-        RobotParams.set_logging_formatter("brief_console_formatter")
-
-        self.state_keys = None
-        self.action_keys = None
-
-    def connect(self) -> None:
-        self.is_connected = self.startup()
-        if not self.is_connected:
-            print("Another process is already using Stretch. Try running 'stretch_free_robot_process.py'")
-            raise ConnectionError()
-
-        for name in self.cameras:
-            self.cameras[name].connect()
-            self.is_connected = self.is_connected and self.cameras[name].is_connected
-
-        if not self.is_connected:
-            print("Could not connect to the cameras, check that all cameras are plugged-in.")
-            raise ConnectionError()
-
-        self.run_calibration()
-
-    def run_calibration(self) -> None:
-        if not self.is_homed():
-            self.home()
-
-    def teleop_step(
-        self, record_data=False
-    ) -> None | tuple[dict[str, torch.Tensor], dict[str, torch.Tensor]]:
-        # TODO(aliberts): return ndarrays instead of torch.Tensors
-        if not self.is_connected:
-            raise ConnectionError()
-
-        if self.teleop is None:
-            self.teleop = GamePadTeleop(robot_instance=False)
-            self.teleop.startup(robot=self)
-
-        before_read_t = time.perf_counter()
-        state = self.get_state()
-        action = self.teleop.gamepad_controller.get_state()
-        self.logs["read_pos_dt_s"] = time.perf_counter() - before_read_t
-
-        before_write_t = time.perf_counter()
-        self.teleop.do_motion(robot=self)
-        self.push_command()
-        self.logs["write_pos_dt_s"] = time.perf_counter() - before_write_t
-
-        if self.state_keys is None:
-            self.state_keys = list(state)
-
-        if not record_data:
-            return
-
-        state = torch.as_tensor(list(state.values()))
-        action = torch.as_tensor(list(action.values()))
-
-        # Capture images from cameras
-        images = {}
-        for name in self.cameras:
-            before_camread_t = time.perf_counter()
-            images[name] = self.cameras[name].async_read()
-            images[name] = torch.from_numpy(images[name])
-            self.logs[f"read_camera_{name}_dt_s"] = self.cameras[name].logs["delta_timestamp_s"]
-            self.logs[f"async_read_camera_{name}_dt_s"] = time.perf_counter() - before_camread_t
-
-        # Populate output dictionaries
-        obs_dict, action_dict = {}, {}
-        obs_dict["observation.state"] = state
-        action_dict["action"] = action
-        for name in self.cameras:
-            obs_dict[f"observation.images.{name}"] = images[name]
-
-        return obs_dict, action_dict
-
-    def get_state(self) -> dict:
-        status = self.get_status()
-        return {
-            "head_pan.pos": status["head"]["head_pan"]["pos"],
-            "head_tilt.pos": status["head"]["head_tilt"]["pos"],
-            "lift.pos": status["lift"]["pos"],
-            "arm.pos": status["arm"]["pos"],
-            "wrist_pitch.pos": status["end_of_arm"]["wrist_pitch"]["pos"],
-            "wrist_roll.pos": status["end_of_arm"]["wrist_roll"]["pos"],
-            "wrist_yaw.pos": status["end_of_arm"]["wrist_yaw"]["pos"],
-            "gripper.pos": status["end_of_arm"]["stretch_gripper"]["pos"],
-            "base_x.vel": status["base"]["x_vel"],
-            "base_y.vel": status["base"]["y_vel"],
-            "base_theta.vel": status["base"]["theta_vel"],
-        }
-
-    def capture_observation(self) -> dict:
-        # TODO(aliberts): return ndarrays instead of torch.Tensors
-        before_read_t = time.perf_counter()
-        state = self.get_state()
-        self.logs["read_pos_dt_s"] = time.perf_counter() - before_read_t
-
-        if self.state_keys is None:
-            self.state_keys = list(state)
-
-        state = torch.as_tensor(list(state.values()))
-
-        # Capture images from cameras
-        images = {}
-        for name in self.cameras:
-            before_camread_t = time.perf_counter()
-            images[name] = self.cameras[name].async_read()
-            images[name] = torch.from_numpy(images[name])
-            self.logs[f"read_camera_{name}_dt_s"] = self.cameras[name].logs["delta_timestamp_s"]
-            self.logs[f"async_read_camera_{name}_dt_s"] = time.perf_counter() - before_camread_t
-
-        # Populate output dictionaries
-        obs_dict = {}
-        obs_dict["observation.state"] = state
-        for name in self.cameras:
-            obs_dict[f"observation.images.{name}"] = images[name]
-
-        return obs_dict
-
-    def send_action(self, action: torch.Tensor) -> torch.Tensor:
-        # TODO(aliberts): return ndarrays instead of torch.Tensors
-        if not self.is_connected:
-            raise ConnectionError()
-
-        if self.teleop is None:
-            self.teleop = GamePadTeleop(robot_instance=False)
-            self.teleop.startup(robot=self)
-
-        if self.action_keys is None:
-            dummy_action = self.teleop.gamepad_controller.get_state()
-            self.action_keys = list(dummy_action.keys())
-
-        action_dict = dict(zip(self.action_keys, action.tolist(), strict=True))
-
-        before_write_t = time.perf_counter()
-        self.teleop.do_motion(state=action_dict, robot=self)
-        self.push_command()
-        self.logs["write_pos_dt_s"] = time.perf_counter() - before_write_t
-
-        # TODO(aliberts): return action_sent when motion is limited
-        return action
-
-    def print_logs(self) -> None:
-        pass
-        # TODO(aliberts): move robot-specific logs logic here
-
-    def teleop_safety_stop(self) -> None:
-        if self.teleop is not None:
-            self.teleop._safety_stop(robot=self)
-
-    def disconnect(self) -> None:
-        self.stop()
-        if self.teleop is not None:
-            self.teleop.gamepad_controller.stop()
-            self.teleop.stop()
-
-        if len(self.cameras) > 0:
-            for cam in self.cameras.values():
-                cam.disconnect()
-
-        self.is_connected = False
-
-    def __del__(self):
-        self.disconnect()

lerobot/common/robot_devices/robots/utils.py

@@ -1,89 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from typing import Protocol
-
-from lerobot.common.robot_devices.robots.configs import (
-    AlohaRobotConfig,
-    KochBimanualRobotConfig,
-    KochRobotConfig,
-    LeKiwiRobotConfig,
-    ManipulatorRobotConfig,
-    MossRobotConfig,
-    RobotConfig,
-    So100RobotConfig,
-    So101RobotConfig,
-    StretchRobotConfig,
-)
-
-
-def get_arm_id(name, arm_type):
-    """Returns the string identifier of a robot arm. For instance, for a bimanual manipulator
-    like Aloha, it could be left_follower, right_follower, left_leader, or right_leader.
-    """
-    return f"{name}_{arm_type}"
-
-
-class Robot(Protocol):
-    # TODO(rcadene, aliberts): Add unit test checking the protocol is implemented in the corresponding classes
-    robot_type: str
-    features: dict
-
-    def connect(self): ...
-    def run_calibration(self): ...
-    def teleop_step(self, record_data=False): ...
-    def capture_observation(self): ...
-    def send_action(self, action): ...
-    def disconnect(self): ...
-
-
-def make_robot_config(robot_type: str, **kwargs) -> RobotConfig:
-    if robot_type == "aloha":
-        return AlohaRobotConfig(**kwargs)
-    elif robot_type == "koch":
-        return KochRobotConfig(**kwargs)
-    elif robot_type == "koch_bimanual":
-        return KochBimanualRobotConfig(**kwargs)
-    elif robot_type == "moss":
-        return MossRobotConfig(**kwargs)
-    elif robot_type == "so100":
-        return So100RobotConfig(**kwargs)
-    elif robot_type == "so101":
-        return So101RobotConfig(**kwargs)
-    elif robot_type == "stretch":
-        return StretchRobotConfig(**kwargs)
-    elif robot_type == "lekiwi":
-        return LeKiwiRobotConfig(**kwargs)
-    else:
-        raise ValueError(f"Robot type '{robot_type}' is not available.")
-
-
-def make_robot_from_config(config: RobotConfig):
-    if isinstance(config, ManipulatorRobotConfig):
-        from lerobot.common.robot_devices.robots.manipulator import ManipulatorRobot
-
-        return ManipulatorRobot(config)
-    elif isinstance(config, LeKiwiRobotConfig):
-        from lerobot.common.robot_devices.robots.mobile_manipulator import MobileManipulator
-
-        return MobileManipulator(config)
-    else:
-        from lerobot.common.robot_devices.robots.stretch import StretchRobot
-
-        return StretchRobot(config)
-
-
-def make_robot(robot_type: str, **kwargs) -> Robot:
-    config = make_robot_config(robot_type, **kwargs)
-    return make_robot_from_config(config)

lerobot/scripts/configure_motor.py

@@ -1,176 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""
-This script configure a single motor at a time to a given ID and baudrate.
-
-Example of usage:
-```bash
-python lerobot/scripts/configure_motor.py \
-  --port /dev/tty.usbmodem585A0080521 \
-  --brand feetech \
-  --model sts3215 \
-  --baudrate 1000000 \
-  --ID 1
-```
-"""
-
-import argparse
-import time
-
-
-def get_motor_bus_cls(brand: str) -> tuple:
-    if brand == "feetech":
-        from lerobot.common.robot_devices.motors.configs import FeetechMotorsBusConfig
-        from lerobot.common.robot_devices.motors.feetech import (
-            MODEL_BAUDRATE_TABLE,
-            SCS_SERIES_BAUDRATE_TABLE,
-            FeetechMotorsBus,
-        )
-
-        return FeetechMotorsBusConfig, FeetechMotorsBus, MODEL_BAUDRATE_TABLE, SCS_SERIES_BAUDRATE_TABLE
-
-    elif brand == "dynamixel":
-        from lerobot.common.robot_devices.motors.configs import DynamixelMotorsBusConfig
-        from lerobot.common.robot_devices.motors.dynamixel import (
-            MODEL_BAUDRATE_TABLE,
-            X_SERIES_BAUDRATE_TABLE,
-            DynamixelMotorsBus,
-        )
-
-        return DynamixelMotorsBusConfig, DynamixelMotorsBus, MODEL_BAUDRATE_TABLE, X_SERIES_BAUDRATE_TABLE
-
-    else:
-        raise ValueError(
-            f"Currently we do not support this motor brand: {brand}. We currently support feetech and dynamixel motors."
-        )
-
-
-def configure_motor(port, brand, model, motor_idx_des, baudrate_des):
-    motor_bus_config_cls, motor_bus_cls, model_baudrate_table, series_baudrate_table = get_motor_bus_cls(
-        brand
-    )
-
-    # Check if the provided model exists in the model_baud_rate_table
-    if model not in model_baudrate_table:
-        raise ValueError(
-            f"Invalid model '{model}' for brand '{brand}'. Supported models: {list(model_baudrate_table.keys())}"
-        )
-
-    # Setup motor names, indices, and models
-    motor_name = "motor"
-    motor_index_arbitrary = motor_idx_des  # Use the motor ID passed via argument
-    motor_model = model  # Use the motor model passed via argument
-
-    config = motor_bus_config_cls(port=port, motors={motor_name: (motor_index_arbitrary, motor_model)})
-
-    # Initialize the MotorBus with the correct port and motor configurations
-    motor_bus = motor_bus_cls(config=config)
-
-    # Try to connect to the motor bus and handle any connection-specific errors
-    try:
-        motor_bus.connect()
-        print(f"Connected on port {motor_bus.port}")
-    except OSError as e:
-        print(f"Error occurred when connecting to the motor bus: {e}")
-        return
-
-    # Motor bus is connected, proceed with the rest of the operations
-    try:
-        print("Scanning all baudrates and motor indices")
-        all_baudrates = set(series_baudrate_table.values())
-        motor_index = -1  # Set the motor index to an out-of-range value.
-
-        for baudrate in all_baudrates:
-            motor_bus.set_bus_baudrate(baudrate)
-            present_ids = motor_bus.find_motor_indices(list(range(1, 10)))
-            if len(present_ids) > 1:
-                raise ValueError(
-                    "Error: More than one motor ID detected. This script is designed to only handle one motor at a time. Please disconnect all but one motor."
-                )
-
-            if len(present_ids) == 1:
-                if motor_index != -1:
-                    raise ValueError(
-                        "Error: More than one motor ID detected. This script is designed to only handle one motor at a time. Please disconnect all but one motor."
-                    )
-                motor_index = present_ids[0]
-                break
-
-        if motor_index == -1:
-            raise ValueError("No motors detected. Please ensure you have one motor connected.")
-
-        print(f"Motor index found at: {motor_index}")
-
-        if brand == "feetech":
-            # Allows ID and BAUDRATE to be written in memory
-            motor_bus.write_with_motor_ids(motor_bus.motor_models, motor_index, "Lock", 0)
-
-        if baudrate != baudrate_des:
-            print(f"Setting its baudrate to {baudrate_des}")
-            baudrate_idx = list(series_baudrate_table.values()).index(baudrate_des)
-
-            # The write can fail, so we allow retries
-            motor_bus.write_with_motor_ids(motor_bus.motor_models, motor_index, "Baud_Rate", baudrate_idx)
-            time.sleep(0.5)
-            motor_bus.set_bus_baudrate(baudrate_des)
-            present_baudrate_idx = motor_bus.read_with_motor_ids(
-                motor_bus.motor_models, motor_index, "Baud_Rate", num_retry=2
-            )
-
-            if present_baudrate_idx != baudrate_idx:
-                raise OSError("Failed to write baudrate.")
-
-        print(f"Setting its index to desired index {motor_idx_des}")
-        if brand == "feetech":
-            motor_bus.write_with_motor_ids(motor_bus.motor_models, motor_index, "Lock", 0)
-        motor_bus.write_with_motor_ids(motor_bus.motor_models, motor_index, "ID", motor_idx_des)
-
-        present_idx = motor_bus.read_with_motor_ids(motor_bus.motor_models, motor_idx_des, "ID", num_retry=2)
-        if present_idx != motor_idx_des:
-            raise OSError("Failed to write index.")
-
-        if brand == "feetech":
-            # Set Maximum_Acceleration to 254 to speedup acceleration and deceleration of
-            # the motors. Note: this configuration is not in the official STS3215 Memory Table
-            motor_bus.write("Lock", 0)
-            motor_bus.write("Maximum_Acceleration", 254)
-
-            motor_bus.write("Goal_Position", 2048)
-            time.sleep(4)
-            print("Present Position", motor_bus.read("Present_Position"))
-
-            motor_bus.write("Offset", 0)
-            time.sleep(4)
-            print("Offset", motor_bus.read("Offset"))
-
-    except Exception as e:
-        print(f"Error occurred during motor configuration: {e}")
-
-    finally:
-        motor_bus.disconnect()
-        print("Disconnected from motor bus.")
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--port", type=str, required=True, help="Motors bus port (e.g. dynamixel,feetech)")
-    parser.add_argument("--brand", type=str, required=True, help="Motor brand (e.g. dynamixel,feetech)")
-    parser.add_argument("--model", type=str, required=True, help="Motor model (e.g. xl330-m077,sts3215)")
-    parser.add_argument("--ID", type=int, required=True, help="Desired ID of the current motor (e.g. 1,2,3)")
-    parser.add_argument(
-        "--baudrate", type=int, default=1000000, help="Desired baudrate for the motor (default: 1000000)"
-    )
-    args = parser.parse_args()
-
-    configure_motor(args.port, args.brand, args.model, args.ID, args.baudrate)

lerobot/scripts/control_robot.py

@@ -1,437 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""
-Utilities to control a robot.
-
-Useful to record a dataset, replay a recorded episode, run the policy on your robot
-and record an evaluation dataset, and to recalibrate your robot if needed.
-
-Examples of usage:
-
-- Recalibrate your robot:
-```bash
-python lerobot/scripts/control_robot.py \
-    --robot.type=so100 \
-    --control.type=calibrate
-```
-
-- Unlimited teleoperation at highest frequency (~200 Hz is expected), to exit with CTRL+C:
-```bash
-python lerobot/scripts/control_robot.py \
-    --robot.type=so100 \
-    --robot.cameras='{}' \
-    --control.type=teleoperate
-
-# Add the cameras from the robot definition to visualize them:
-python lerobot/scripts/control_robot.py \
-    --robot.type=so100 \
-    --control.type=teleoperate
-```
-
-- Unlimited teleoperation at a limited frequency of 30 Hz, to simulate data recording frequency:
-```bash
-python lerobot/scripts/control_robot.py \
-    --robot.type=so100 \
-    --control.type=teleoperate \
-    --control.fps=30
-```
-
-- Record one episode in order to test replay:
-```bash
-python lerobot/scripts/control_robot.py \
-    --robot.type=so100 \
-    --control.type=record \
-    --control.fps=30 \
-    --control.single_task="Grasp a lego block and put it in the bin." \
-    --control.repo_id=$USER/koch_test \
-    --control.num_episodes=1 \
-    --control.push_to_hub=True
-```
-
-- Visualize dataset:
-```bash
-python lerobot/scripts/visualize_dataset.py \
-    --repo-id $USER/koch_test \
-    --episode-index 0
-```
-
-- Replay this test episode:
-```bash
-python lerobot/scripts/control_robot.py replay \
-    --robot.type=so100 \
-    --control.type=replay \
-    --control.fps=30 \
-    --control.repo_id=$USER/koch_test \
-    --control.episode=0
-```
-
-- Record a full dataset in order to train a policy, with 2 seconds of warmup,
-30 seconds of recording for each episode, and 10 seconds to reset the environment in between episodes:
-```bash
-python lerobot/scripts/control_robot.py record \
-    --robot.type=so100 \
-    --control.type=record \
-    --control.fps 30 \
-    --control.repo_id=$USER/koch_pick_place_lego \
-    --control.num_episodes=50 \
-    --control.warmup_time_s=2 \
-    --control.episode_time_s=30 \
-    --control.reset_time_s=10
-```
-
-- For remote controlled robots like LeKiwi, run this script on the robot edge device (e.g. RaspBerryPi):
-```bash
-python lerobot/scripts/control_robot.py \
-  --robot.type=lekiwi \
-  --control.type=remote_robot
-```
-
-**NOTE**: You can use your keyboard to control data recording flow.
-- Tap right arrow key '->' to early exit while recording an episode and go to resseting the environment.
-- Tap right arrow key '->' to early exit while resetting the environment and got to recording the next episode.
-- Tap left arrow key '<-' to early exit and re-record the current episode.
-- Tap escape key 'esc' to stop the data recording.
-This might require a sudo permission to allow your terminal to monitor keyboard events.
-
-**NOTE**: You can resume/continue data recording by running the same data recording command and adding `--control.resume=true`.
-
-- Train on this dataset with the ACT policy:
-```bash
-python lerobot/scripts/train.py \
-  --dataset.repo_id=${HF_USER}/koch_pick_place_lego \
-  --policy.type=act \
-  --output_dir=outputs/train/act_koch_pick_place_lego \
-  --job_name=act_koch_pick_place_lego \
-  --device=cuda \
-  --wandb.enable=true
-```
-
-- Run the pretrained policy on the robot:
-```bash
-python lerobot/scripts/control_robot.py \
-    --robot.type=so100 \
-    --control.type=record \
-    --control.fps=30 \
-    --control.single_task="Grasp a lego block and put it in the bin." \
-    --control.repo_id=$USER/eval_act_koch_pick_place_lego \
-    --control.num_episodes=10 \
-    --control.warmup_time_s=2 \
-    --control.episode_time_s=30 \
-    --control.reset_time_s=10 \
-    --control.push_to_hub=true \
-    --control.policy.path=outputs/train/act_koch_pick_place_lego/checkpoints/080000/pretrained_model
-```
-"""
-
-import logging
-import os
-import time
-from dataclasses import asdict
-from pprint import pformat
-
-import rerun as rr
-
-# from safetensors.torch import load_file, save_file
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
-from lerobot.common.policies.factory import make_policy
-from lerobot.common.robot_devices.control_configs import (
-    CalibrateControlConfig,
-    ControlConfig,
-    ControlPipelineConfig,
-    RecordControlConfig,
-    RemoteRobotConfig,
-    ReplayControlConfig,
-    TeleoperateControlConfig,
-)
-from lerobot.common.robot_devices.control_utils import (
-    control_loop,
-    init_keyboard_listener,
-    is_headless,
-    log_control_info,
-    record_episode,
-    reset_environment,
-    sanity_check_dataset_name,
-    sanity_check_dataset_robot_compatibility,
-    stop_recording,
-    warmup_record,
-)
-from lerobot.common.robot_devices.robots.utils import Robot, make_robot_from_config
-from lerobot.common.robot_devices.utils import busy_wait, safe_disconnect
-from lerobot.common.utils.utils import has_method, init_logging, log_say
-from lerobot.configs import parser
-
-########################################################################################
-# Control modes
-########################################################################################
-
-
-@safe_disconnect
-def calibrate(robot: Robot, cfg: CalibrateControlConfig):
-    # TODO(aliberts): move this code in robots' classes
-    if robot.robot_type.startswith("stretch"):
-        if not robot.is_connected:
-            robot.connect()
-        if not robot.is_homed():
-            robot.home()
-        return
-
-    arms = robot.available_arms if cfg.arms is None else cfg.arms
-    unknown_arms = [arm_id for arm_id in arms if arm_id not in robot.available_arms]
-    available_arms_str = " ".join(robot.available_arms)
-    unknown_arms_str = " ".join(unknown_arms)
-
-    if arms is None or len(arms) == 0:
-        raise ValueError(
-            "No arm provided. Use `--arms` as argument with one or more available arms.\n"
-            f"For instance, to recalibrate all arms add: `--arms {available_arms_str}`"
-        )
-
-    if len(unknown_arms) > 0:
-        raise ValueError(
-            f"Unknown arms provided ('{unknown_arms_str}'). Available arms are `{available_arms_str}`."
-        )
-
-    for arm_id in arms:
-        arm_calib_path = robot.calibration_dir / f"{arm_id}.json"
-        if arm_calib_path.exists():
-            print(f"Removing '{arm_calib_path}'")
-            arm_calib_path.unlink()
-        else:
-            print(f"Calibration file not found '{arm_calib_path}'")
-
-    if robot.is_connected:
-        robot.disconnect()
-
-    if robot.robot_type.startswith("lekiwi") and "main_follower" in arms:
-        print("Calibrating only the lekiwi follower arm 'main_follower'...")
-        robot.calibrate_follower()
-        return
-
-    if robot.robot_type.startswith("lekiwi") and "main_leader" in arms:
-        print("Calibrating only the lekiwi leader arm 'main_leader'...")
-        robot.calibrate_leader()
-        return
-
-    # Calling `connect` automatically runs calibration
-    # when the calibration file is missing
-    robot.connect()
-    robot.disconnect()
-    print("Calibration is done! You can now teleoperate and record datasets!")
-
-
-@safe_disconnect
-def teleoperate(robot: Robot, cfg: TeleoperateControlConfig):
-    control_loop(
-        robot,
-        control_time_s=cfg.teleop_time_s,
-        fps=cfg.fps,
-        teleoperate=True,
-        display_data=cfg.display_data,
-    )
-
-
-@safe_disconnect
-def record(
-    robot: Robot,
-    cfg: RecordControlConfig,
-) -> LeRobotDataset:
-    # TODO(rcadene): Add option to record logs
-    if cfg.resume:
-        dataset = LeRobotDataset(
-            cfg.repo_id,
-            root=cfg.root,
-        )
-        if len(robot.cameras) > 0:
-            dataset.start_image_writer(
-                num_processes=cfg.num_image_writer_processes,
-                num_threads=cfg.num_image_writer_threads_per_camera * len(robot.cameras),
-            )
-        sanity_check_dataset_robot_compatibility(dataset, robot, cfg.fps, cfg.video)
-    else:
-        # Create empty dataset or load existing saved episodes
-        sanity_check_dataset_name(cfg.repo_id, cfg.policy)
-        dataset = LeRobotDataset.create(
-            cfg.repo_id,
-            cfg.fps,
-            root=cfg.root,
-            robot=robot,
-            use_videos=cfg.video,
-            image_writer_processes=cfg.num_image_writer_processes,
-            image_writer_threads=cfg.num_image_writer_threads_per_camera * len(robot.cameras),
-        )
-
-    # Load pretrained policy
-    policy = None if cfg.policy is None else make_policy(cfg.policy, ds_meta=dataset.meta)
-
-    if not robot.is_connected:
-        robot.connect()
-
-    listener, events = init_keyboard_listener()
-
-    # Execute a few seconds without recording to:
-    # 1. teleoperate the robot to move it in starting position if no policy provided,
-    # 2. give times to the robot devices to connect and start synchronizing,
-    # 3. place the cameras windows on screen
-    enable_teleoperation = policy is None
-    log_say("Warmup record", cfg.play_sounds)
-    warmup_record(robot, events, enable_teleoperation, cfg.warmup_time_s, cfg.display_data, cfg.fps)
-
-    if has_method(robot, "teleop_safety_stop"):
-        robot.teleop_safety_stop()
-
-    recorded_episodes = 0
-    while True:
-        if recorded_episodes >= cfg.num_episodes:
-            break
-
-        log_say(f"Recording episode {dataset.num_episodes}", cfg.play_sounds)
-        record_episode(
-            robot=robot,
-            dataset=dataset,
-            events=events,
-            episode_time_s=cfg.episode_time_s,
-            display_data=cfg.display_data,
-            policy=policy,
-            fps=cfg.fps,
-            single_task=cfg.single_task,
-        )
-
-        # Execute a few seconds without recording to give time to manually reset the environment
-        # Current code logic doesn't allow to teleoperate during this time.
-        # TODO(rcadene): add an option to enable teleoperation during reset
-        # Skip reset for the last episode to be recorded
-        if not events["stop_recording"] and (
-            (recorded_episodes < cfg.num_episodes - 1) or events["rerecord_episode"]
-        ):
-            log_say("Reset the environment", cfg.play_sounds)
-            reset_environment(robot, events, cfg.reset_time_s, cfg.fps)
-
-        if events["rerecord_episode"]:
-            log_say("Re-record episode", cfg.play_sounds)
-            events["rerecord_episode"] = False
-            events["exit_early"] = False
-            dataset.clear_episode_buffer()
-            continue
-
-        dataset.save_episode()
-        recorded_episodes += 1
-
-        if events["stop_recording"]:
-            break
-
-    log_say("Stop recording", cfg.play_sounds, blocking=True)
-    stop_recording(robot, listener, cfg.display_data)
-
-    if cfg.push_to_hub:
-        dataset.push_to_hub(tags=cfg.tags, private=cfg.private)
-
-    log_say("Exiting", cfg.play_sounds)
-    return dataset
-
-
-@safe_disconnect
-def replay(
-    robot: Robot,
-    cfg: ReplayControlConfig,
-):
-    # TODO(rcadene, aliberts): refactor with control_loop, once `dataset` is an instance of LeRobotDataset
-    # TODO(rcadene): Add option to record logs
-
-    dataset = LeRobotDataset(cfg.repo_id, root=cfg.root, episodes=[cfg.episode])
-    actions = dataset.hf_dataset.select_columns("action")
-
-    if not robot.is_connected:
-        robot.connect()
-
-    log_say("Replaying episode", cfg.play_sounds, blocking=True)
-    for idx in range(dataset.num_frames):
-        start_episode_t = time.perf_counter()
-
-        action = actions[idx]["action"]
-        robot.send_action(action)
-
-        dt_s = time.perf_counter() - start_episode_t
-        busy_wait(1 / cfg.fps - dt_s)
-
-        dt_s = time.perf_counter() - start_episode_t
-        log_control_info(robot, dt_s, fps=cfg.fps)
-
-
-def _init_rerun(control_config: ControlConfig, session_name: str = "lerobot_control_loop") -> None:
-    """Initializes the Rerun SDK for visualizing the control loop.
-
-    Args:
-        control_config: Configuration determining data display and robot type.
-        session_name: Rerun session name. Defaults to "lerobot_control_loop".
-
-    Raises:
-        ValueError: If viewer IP is missing for non-remote configurations with display enabled.
-    """
-    if (control_config.display_data and not is_headless()) or (
-        control_config.display_data and isinstance(control_config, RemoteRobotConfig)
-    ):
-        # Configure Rerun flush batch size default to 8KB if not set
-        batch_size = os.getenv("RERUN_FLUSH_NUM_BYTES", "8000")
-        os.environ["RERUN_FLUSH_NUM_BYTES"] = batch_size
-
-        # Initialize Rerun based on configuration
-        rr.init(session_name)
-        if isinstance(control_config, RemoteRobotConfig):
-            viewer_ip = control_config.viewer_ip
-            viewer_port = control_config.viewer_port
-            if not viewer_ip or not viewer_port:
-                raise ValueError(
-                    "Viewer IP & Port are required for remote config. Set via config file/CLI or disable control_config.display_data."
-                )
-            logging.info(f"Connecting to viewer at {viewer_ip}:{viewer_port}")
-            rr.connect_tcp(f"{viewer_ip}:{viewer_port}")
-        else:
-            # Get memory limit for rerun viewer parameters
-            memory_limit = os.getenv("LEROBOT_RERUN_MEMORY_LIMIT", "10%")
-            rr.spawn(memory_limit=memory_limit)
-
-
-@parser.wrap()
-def control_robot(cfg: ControlPipelineConfig):
-    init_logging()
-    logging.info(pformat(asdict(cfg)))
-
-    robot = make_robot_from_config(cfg.robot)
-
-    # TODO(Steven): Blueprint for fixed window size
-
-    if isinstance(cfg.control, CalibrateControlConfig):
-        calibrate(robot, cfg.control)
-    elif isinstance(cfg.control, TeleoperateControlConfig):
-        _init_rerun(control_config=cfg.control, session_name="lerobot_control_loop_teleop")
-        teleoperate(robot, cfg.control)
-    elif isinstance(cfg.control, RecordControlConfig):
-        _init_rerun(control_config=cfg.control, session_name="lerobot_control_loop_record")
-        record(robot, cfg.control)
-    elif isinstance(cfg.control, ReplayControlConfig):
-        replay(robot, cfg.control)
-    elif isinstance(cfg.control, RemoteRobotConfig):
-        from lerobot.common.robot_devices.robots.lekiwi_remote import run_lekiwi
-
-        _init_rerun(control_config=cfg.control, session_name="lerobot_control_loop_remote")
-        run_lekiwi(cfg.robot)
-
-    if robot.is_connected:
-        # Disconnect manually to avoid a "Core dump" during process
-        # termination due to camera threads not properly exiting.
-        robot.disconnect()
-
-
-if __name__ == "__main__":
-    control_robot()

lerobot/scripts/control_sim_robot.py

@@ -1,561 +0,0 @@
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""
-Utilities to control a robot in simulation.
-
-Useful to record a dataset, replay a recorded episode and record an evaluation dataset.
-
-Examples of usage:
-
-
-- Unlimited teleoperation at a limited frequency of 30 Hz, to simulate data recording frequency.
-  You can modify this value depending on how fast your simulation can run:
-```bash
-python lerobot/scripts/control_robot.py teleoperate \
-    --fps 30 \
-    --robot-path lerobot/configs/robot/your_robot_config.yaml \
-    --sim-config lerobot/configs/env/your_sim_config.yaml
-```
-
-- Record one episode in order to test replay:
-```bash
-python lerobot/scripts/control_sim_robot.py record \
-    --robot-path lerobot/configs/robot/your_robot_config.yaml \
-    --sim-config lerobot/configs/env/your_sim_config.yaml \
-    --fps 30 \
-    --repo-id $USER/robot_sim_test \
-    --num-episodes 1 \
-    --run-compute-stats 0
-```
-
-Enable the --push-to-hub 1 to push the recorded dataset to the huggingface hub.
-
-- Visualize dataset:
-```bash
-python lerobot/scripts/visualize_dataset.py \
-    --repo-id $USER/robot_sim_test \
-    --episode-index 0
-```
-
-- Replay a sequence of test episodes:
-```bash
-python lerobot/scripts/control_sim_robot.py replay \
-    --robot-path lerobot/configs/robot/your_robot_config.yaml \
-    --sim-config lerobot/configs/env/your_sim_config.yaml \
-    --fps 30 \
-    --repo-id $USER/robot_sim_test \
-    --episode 0
-```
-Note: The seed is saved, therefore, during replay we can load the same environment state as the one during collection.
-
-- Record a full dataset in order to train a policy,
-30 seconds of recording for each episode, and 10 seconds to reset the environment in between episodes:
-```bash
-python lerobot/scripts/control_sim_robot.py record \
-    --robot-path lerobot/configs/robot/your_robot_config.yaml \
-    --sim-config lerobot/configs/env/your_sim_config.yaml \
-    --fps 30 \
-    --repo-id $USER/robot_sim_test \
-    --num-episodes 50 \
-    --episode-time-s 30 \
-```
-
-**NOTE**: You can use your keyboard to control data recording flow.
-- Tap right arrow key '->' to early exit while recording an episode and go to resetting the environment.
-- Tap right arrow key '->' to early exit while resetting the environment and got to recording the next episode.
-- Tap left arrow key '<-' to early exit and re-record the current episode.
-- Tap escape key 'esc' to stop the data recording.
-This might require a sudo permission to allow your terminal to monitor keyboard events.
-
-**NOTE**: You can resume/continue data recording by running the same data recording command twice.
-"""
-
-import argparse
-import importlib
-import logging
-import time
-from pathlib import Path
-
-import cv2
-import gymnasium as gym
-import numpy as np
-import torch
-
-from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
-from lerobot.common.robot_devices.control_utils import (
-    init_keyboard_listener,
-    init_policy,
-    is_headless,
-    log_control_info,
-    predict_action,
-    sanity_check_dataset_name,
-    sanity_check_dataset_robot_compatibility,
-    stop_recording,
-)
-from lerobot.common.robot_devices.robots.utils import Robot, make_robot
-from lerobot.common.robot_devices.utils import busy_wait
-from lerobot.common.utils.utils import init_hydra_config, init_logging, log_say
-
-raise NotImplementedError("This script is currently deactivated")
-
-DEFAULT_FEATURES = {
-    "next.reward": {
-        "dtype": "float32",
-        "shape": (1,),
-        "names": None,
-    },
-    "next.success": {
-        "dtype": "bool",
-        "shape": (1,),
-        "names": None,
-    },
-    "seed": {
-        "dtype": "int64",
-        "shape": (1,),
-        "names": None,
-    },
-    "timestamp": {
-        "dtype": "float32",
-        "shape": (1,),
-        "names": None,
-    },
-}
-
-
-########################################################################################
-# Utilities
-########################################################################################
-def none_or_int(value):
-    if value == "None":
-        return None
-    return int(value)
-
-
-def init_sim_calibration(robot, cfg):
-    # Constants necessary for transforming the joint pos of the real robot to the sim
-    # depending on the robot description used in that sim.
-    start_pos = np.array(robot.leader_arms.main.calibration["start_pos"])
-    axis_directions = np.array(cfg.get("axis_directions", [1]))
-    offsets = np.array(cfg.get("offsets", [0])) * np.pi
-
-    return {"start_pos": start_pos, "axis_directions": axis_directions, "offsets": offsets}
-
-
-def real_positions_to_sim(real_positions, axis_directions, start_pos, offsets):
-    """Counts - starting position -> radians -> align axes -> offset"""
-    return axis_directions * (real_positions - start_pos) * 2.0 * np.pi / 4096 + offsets
-
-
-########################################################################################
-# Control modes
-########################################################################################
-
-
-def teleoperate(env, robot: Robot, process_action_fn, teleop_time_s=None):
-    env = env()
-    env.reset()
-    start_teleop_t = time.perf_counter()
-    while True:
-        leader_pos = robot.leader_arms.main.read("Present_Position")
-        action = process_action_fn(leader_pos)
-        env.step(np.expand_dims(action, 0))
-        if teleop_time_s is not None and time.perf_counter() - start_teleop_t > teleop_time_s:
-            print("Teleoperation processes finished.")
-            break
-
-
-def record(
-    env,
-    robot: Robot,
-    process_action_from_leader,
-    root: Path,
-    repo_id: str,
-    task: str,
-    fps: int | None = None,
-    tags: list[str] | None = None,
-    pretrained_policy_name_or_path: str = None,
-    policy_overrides: bool | None = None,
-    episode_time_s: int = 30,
-    num_episodes: int = 50,
-    video: bool = True,
-    push_to_hub: bool = True,
-    num_image_writer_processes: int = 0,
-    num_image_writer_threads_per_camera: int = 4,
-    display_cameras: bool = False,
-    play_sounds: bool = True,
-    resume: bool = False,
-    local_files_only: bool = False,
-    run_compute_stats: bool = True,
-) -> LeRobotDataset:
-    # Load pretrained policy
-    policy = None
-    if pretrained_policy_name_or_path is not None:
-        policy, policy_fps, device, use_amp = init_policy(pretrained_policy_name_or_path, policy_overrides)
-
-        if fps is None:
-            fps = policy_fps
-            logging.warning(f"No fps provided, so using the fps from policy config ({policy_fps}).")
-
-    if policy is None and process_action_from_leader is None:
-        raise ValueError("Either policy or process_action_fn has to be set to enable control in sim.")
-
-    # initialize listener before sim env
-    listener, events = init_keyboard_listener()
-
-    # create sim env
-    env = env()
-
-    # Create empty dataset or load existing saved episodes
-    num_cameras = sum([1 if "image" in key else 0 for key in env.observation_space])
-
-    # get image keys
-    image_keys = [key for key in env.observation_space if "image" in key]
-    state_keys_dict = env_cfg.state_keys
-
-    if resume:
-        dataset = LeRobotDataset(
-            repo_id,
-            root=root,
-            local_files_only=local_files_only,
-        )
-        dataset.start_image_writer(
-            num_processes=num_image_writer_processes,
-            num_threads=num_image_writer_threads_per_camera * num_cameras,
-        )
-        sanity_check_dataset_robot_compatibility(dataset, robot, fps, video)
-    else:
-        features = DEFAULT_FEATURES
-        # add image keys to features
-        for key in image_keys:
-            shape = env.observation_space[key].shape
-            if not key.startswith("observation.image."):
-                key = "observation.image." + key
-            features[key] = {"dtype": "video", "names": ["channels", "height", "width"], "shape": shape}
-
-        for key, obs_key in state_keys_dict.items():
-            features[key] = {
-                "dtype": "float32",
-                "names": None,
-                "shape": env.observation_space[obs_key].shape,
-            }
-
-        features["action"] = {"dtype": "float32", "shape": env.action_space.shape, "names": None}
-
-        # Create empty dataset or load existing saved episodes
-        sanity_check_dataset_name(repo_id, policy)
-        dataset = LeRobotDataset.create(
-            repo_id,
-            fps,
-            root=root,
-            features=features,
-            use_videos=video,
-            image_writer_processes=num_image_writer_processes,
-            image_writer_threads=num_image_writer_threads_per_camera * num_cameras,
-        )
-
-    recorded_episodes = 0
-    while True:
-        log_say(f"Recording episode {dataset.num_episodes}", play_sounds)
-
-        if events is None:
-            events = {"exit_early": False}
-
-        if episode_time_s is None:
-            episode_time_s = float("inf")
-
-        timestamp = 0
-        start_episode_t = time.perf_counter()
-
-        seed = np.random.randint(0, 1e5)
-        observation, info = env.reset(seed=seed)
-
-        while timestamp < episode_time_s:
-            start_loop_t = time.perf_counter()
-
-            if policy is not None:
-                action = predict_action(observation, policy, device, use_amp)
-            else:
-                leader_pos = robot.leader_arms.main.read("Present_Position")
-                action = process_action_from_leader(leader_pos)
-
-            observation, reward, terminated, _, info = env.step(action)
-
-            success = info.get("is_success", False)
-            env_timestamp = info.get("timestamp", dataset.episode_buffer["size"] / fps)
-
-            frame = {
-                "action": torch.from_numpy(action),
-                "next.reward": reward,
-                "next.success": success,
-                "seed": seed,
-                "timestamp": env_timestamp,
-            }
-
-            for key in image_keys:
-                if not key.startswith("observation.image"):
-                    frame["observation.image." + key] = observation[key]
-                else:
-                    frame[key] = observation[key]
-
-            for key, obs_key in state_keys_dict.items():
-                frame[key] = torch.from_numpy(observation[obs_key])
-
-            dataset.add_frame(frame)
-
-            if display_cameras and not is_headless():
-                for key in image_keys:
-                    cv2.imshow(key, cv2.cvtColor(observation[key], cv2.COLOR_RGB2BGR))
-                cv2.waitKey(1)
-
-            if fps is not None:
-                dt_s = time.perf_counter() - start_loop_t
-                busy_wait(1 / fps - dt_s)
-
-            dt_s = time.perf_counter() - start_loop_t
-            log_control_info(robot, dt_s, fps=fps)
-
-            timestamp = time.perf_counter() - start_episode_t
-            if events["exit_early"] or terminated:
-                events["exit_early"] = False
-                break
-
-        if events["rerecord_episode"]:
-            log_say("Re-record episode", play_sounds)
-            events["rerecord_episode"] = False
-            events["exit_early"] = False
-            dataset.clear_episode_buffer()
-            continue
-
-        dataset.save_episode(task=task)
-        recorded_episodes += 1
-
-        if events["stop_recording"] or recorded_episodes >= num_episodes:
-            break
-        else:
-            logging.info("Waiting for a few seconds before starting next episode recording...")
-            busy_wait(3)
-
-    log_say("Stop recording", play_sounds, blocking=True)
-    stop_recording(robot, listener, display_cameras)
-
-    if run_compute_stats:
-        logging.info("Computing dataset statistics")
-    dataset.consolidate(run_compute_stats)
-
-    if push_to_hub:
-        dataset.push_to_hub(tags=tags)
-
-    log_say("Exiting", play_sounds)
-    return dataset
-
-
-def replay(
-    env, root: Path, repo_id: str, episode: int, fps: int | None = None, local_files_only: bool = True
-):
-    env = env()
-
-    local_dir = Path(root) / repo_id
-    if not local_dir.exists():
-        raise ValueError(local_dir)
-
-    dataset = LeRobotDataset(repo_id, root=root, local_files_only=local_files_only)
-    items = dataset.hf_dataset.select_columns("action")
-    seeds = dataset.hf_dataset.select_columns("seed")["seed"]
-
-    from_idx = dataset.episode_data_index["from"][episode].item()
-    to_idx = dataset.episode_data_index["to"][episode].item()
-    env.reset(seed=seeds[from_idx].item())
-    logging.info("Replaying episode")
-    log_say("Replaying episode", play_sounds=True)
-    for idx in range(from_idx, to_idx):
-        start_episode_t = time.perf_counter()
-        action = items[idx]["action"]
-        env.step(action.unsqueeze(0).numpy())
-        dt_s = time.perf_counter() - start_episode_t
-        busy_wait(1 / fps - dt_s)
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-    subparsers = parser.add_subparsers(dest="mode", required=True)
-
-    # Set common options for all the subparsers
-    base_parser = argparse.ArgumentParser(add_help=False)
-    base_parser.add_argument(
-        "--robot-path",
-        type=str,
-        default="lerobot/configs/robot/koch.yaml",
-        help="Path to robot yaml file used to instantiate the robot using `make_robot` factory function.",
-    )
-
-    base_parser.add_argument(
-        "--sim-config",
-        help="Path to a yaml config you want to use for initializing a sim environment based on gym ",
-    )
-
-    parser_record = subparsers.add_parser("teleoperate", parents=[base_parser])
-
-    parser_record = subparsers.add_parser("record", parents=[base_parser])
-    parser_record.add_argument(
-        "--fps", type=none_or_int, default=None, help="Frames per second (set to None to disable)"
-    )
-    parser_record.add_argument(
-        "--root",
-        type=Path,
-        default=None,
-        help="Root directory where the dataset will be stored locally at '{root}/{repo_id}' (e.g. 'data/hf_username/dataset_name').",
-    )
-    parser_record.add_argument(
-        "--repo-id",
-        type=str,
-        default="lerobot/test",
-        help="Dataset identifier. By convention it should match '{hf_username}/{dataset_name}' (e.g. `lerobot/test`).",
-    )
-    parser_record.add_argument(
-        "--episode-time-s",
-        type=int,
-        default=60,
-        help="Number of seconds for data recording for each episode.",
-    )
-    parser_record.add_argument(
-        "--task",
-        type=str,
-        required=True,
-        help="A description of the task preformed during recording that can be used as a language instruction.",
-    )
-    parser_record.add_argument("--num-episodes", type=int, default=50, help="Number of episodes to record.")
-    parser_record.add_argument(
-        "--run-compute-stats",
-        type=int,
-        default=1,
-        help="By default, run the computation of the data statistics at the end of data collection. Compute intensive and not required to just replay an episode.",
-    )
-    parser_record.add_argument(
-        "--push-to-hub",
-        type=int,
-        default=1,
-        help="Upload dataset to Hugging Face hub.",
-    )
-    parser_record.add_argument(
-        "--tags",
-        type=str,
-        nargs="*",
-        help="Add tags to your dataset on the hub.",
-    )
-    parser_record.add_argument(
-        "--num-image-writer-processes",
-        type=int,
-        default=0,
-        help=(
-            "Number of subprocesses handling the saving of frames as PNG. Set to 0 to use threads only; "
-            "set to ≥1 to use subprocesses, each using threads to write images. The best number of processes "
-            "and threads depends on your system. We recommend 4 threads per camera with 0 processes. "
-            "If fps is unstable, adjust the thread count. If still unstable, try using 1 or more subprocesses."
-        ),
-    )
-    parser_record.add_argument(
-        "--num-image-writer-threads-per-camera",
-        type=int,
-        default=4,
-        help=(
-            "Number of threads writing the frames as png images on disk, per camera. "
-            "Too much threads might cause unstable teleoperation fps due to main thread being blocked. "
-            "Not enough threads might cause low camera fps."
-        ),
-    )
-    parser_record.add_argument(
-        "--display-cameras",
-        type=int,
-        default=0,
-        help="Visualize image observations with opencv.",
-    )
-    parser_record.add_argument(
-        "--resume",
-        type=int,
-        default=0,
-        help="Resume recording on an existing dataset.",
-    )
-    parser_replay = subparsers.add_parser("replay", parents=[base_parser])
-    parser_replay.add_argument(
-        "--fps", type=none_or_int, default=None, help="Frames per second (set to None to disable)"
-    )
-    parser_replay.add_argument(
-        "--root",
-        type=Path,
-        default=None,
-        help="Root directory where the dataset will be stored locally (e.g. 'data/hf_username/dataset_name'). By default, stored in cache folder.",
-    )
-    parser_replay.add_argument(
-        "--repo-id",
-        type=str,
-        default="lerobot/test",
-        help="Dataset identifier. By convention it should match '{hf_username}/{dataset_name}' (e.g. `lerobot/test`).",
-    )
-    parser_replay.add_argument("--episode", type=int, default=0, help="Index of the episodes to replay.")
-
-    args = parser.parse_args()
-
-    init_logging()
-
-    control_mode = args.mode
-    robot_path = args.robot_path
-    env_config_path = args.sim_config
-    kwargs = vars(args)
-    del kwargs["mode"]
-    del kwargs["robot_path"]
-    del kwargs["sim_config"]
-
-    # make gym env
-    env_cfg = init_hydra_config(env_config_path)
-    importlib.import_module(f"gym_{env_cfg.env.type}")
-
-    def env_constructor():
-        return gym.make(env_cfg.env.handle, disable_env_checker=True, **env_cfg.env.gym)
-
-    robot = None
-    process_leader_actions_fn = None
-
-    if control_mode in ["teleoperate", "record"]:
-        # make robot
-        robot_overrides = ["~cameras", "~follower_arms"]
-        # TODO(rcadene): remove
-        robot_cfg = init_hydra_config(robot_path, robot_overrides)
-        robot = make_robot(robot_cfg)
-        robot.connect()
-
-        calib_kwgs = init_sim_calibration(robot, env_cfg.calibration)
-
-        def process_leader_actions_fn(action):
-            return real_positions_to_sim(action, **calib_kwgs)
-
-        robot.leader_arms.main.calibration = None
-
-    if control_mode == "teleoperate":
-        teleoperate(env_constructor, robot, process_leader_actions_fn)
-
-    elif control_mode == "record":
-        record(env_constructor, robot, process_leader_actions_fn, **kwargs)
-
-    elif control_mode == "replay":
-        replay(env_constructor, **kwargs)
-
-    else:
-        raise ValueError(
-            f"Invalid control mode: '{control_mode}', only valid modes are teleoperate, record and replay."
-        )
-
-    if robot and robot.is_connected:
-        # Disconnect manually to avoid a "Core dump" during process
-        # termination due to camera threads not properly exiting.
-        robot.disconnect()

lerobot/scripts/push_pretrained.py

@@ -1,71 +0,0 @@
-#!/usr/bin/env python
-
-# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""
-Once you have trained a policy with our training script (lerobot/scripts/train.py), use this script to push it
-to the hub.
-
-Example:
-
-```bash
-python lerobot/scripts/push_pretrained.py \
-    --pretrained_path=outputs/train/act_aloha_sim_transfer_cube_human/checkpoints/last/pretrained_model \
-    --repo_id=lerobot/act_aloha_sim_transfer_cube_human
-```
-"""
-
-from dataclasses import dataclass
-from pathlib import Path
-
-import draccus
-from huggingface_hub import HfApi
-
-
-@dataclass
-class PushPreTrainedConfig:
-    pretrained_path: Path
-    repo_id: str
-    branch: str | None = None
-    private: bool = False
-    exist_ok: bool = False
-
-
-@draccus.wrap()
-def main(cfg: PushPreTrainedConfig):
-    hub_api = HfApi()
-    hub_api.create_repo(
-        repo_id=cfg.repo_id,
-        private=cfg.private,
-        repo_type="model",
-        exist_ok=cfg.exist_ok,
-    )
-    if cfg.branch:
-        hub_api.create_branch(
-            repo_id=cfg.repo_id,
-            branch=cfg.branch,
-            repo_type="model",
-            exist_ok=cfg.exist_ok,
-        )
-
-    hub_api.upload_folder(
-        repo_id=cfg.repo_id,
-        folder_path=cfg.pretrained_path,
-        repo_type="model",
-        revision=cfg.branch,
-    )
-
-
-if __name__ == "__main__":
-    main()