fix(missing method): adding missing add_microphones_recordings method in LeRobotDataset

* Add option for pi family models to train with relative actions (relative to state)

* formatting

* add recomputation of stats and option to compute delta stats

* normalzie after delta conversion

* only recompute state for stats

* calulate chunk based stats

* sample 100k

* load from parquet

* sample 1m

* stats per chunck

* fix

* use quantiles

* stats for entire dataset

* fix

* max 1m frames

* compute before dist

* fix multi gpu processor bug

* Fix RTC with delta actions and OpenArms motor_type wiring

* feat: align pi0_fast delta actions with pi0/pi05 and add RTC integration tests

- Add delta_exclude_joints and action_feature_names to PI0FastConfig
- Move to_absolute_actions from modeling to processor pipeline for pi0_fast
- Add delta action detection and logging to eval_with_real_robot.py
- Add delta actions documentation to pi0 and pi05 READMEs
- Fix ruff lint issues in test_delta_actions.py
- Add test_rtc_delta_actions.py (24 tests) covering:
  - ActionQueue with delta vs absolute actions
  - RTC denoise step with delta leftovers
  - Full pipeline roundtrip (delta → RTC → absolute)
  - State rebasing approximation bounds
  - Non-delta policy compatibility
  - Multi-chunk consistency

* chore: clean up test comments, add OpenPI attribution, remove debug logging

- Replace decorative comment separators in test files with plain section headers
- Add attribution comments for 1e-6 epsilon in normalize_processor.py (from OpenPI)
- Remove debug logging blocks from lerobot_train.py

* refactor: extract compute_delta_action_stats into compute_stats.py

Move the ~70-line inline delta action stats block from lerobot_train.py
into a dedicated function in compute_stats.py, where all other stats
computation already lives. The training script now calls it in 6 lines.

* refactor: remove unused get_processed_left_over from ActionQueue

This method was never called outside of tests. Leftover actions for RTC
guidance are always retrieved via get_left_over() (delta/original space).

* revert: remove logging-only changes from eval_with_real_robot.py

The delta actions detection helper and log message added no functional
value — the script already handles delta policies correctly via the
processor pipeline.

* refactor: use ACTION/OBS_STATE constants instead of hardcoded strings

Replace hardcoded "action" and "observation.state" with ACTION and
OBS_STATE from utils.constants in compute_stats.py, dataset_tools.py,
and lerobot_train.py.

* style: remove stray blank lines in training loop

* refactor: move delta action stats to preprocessing step, remove on-the-fly computation

- Remove on-the-fly compute_delta_action_stats from lerobot_train.py
- Rewrite recompute_stats to delegate action stats to compute_delta_action_stats
  (chunk-based sampling matching what the model sees during training)
- Add chunk_size parameter to recompute_stats for delta action computation
- Add delta actions documentation to pi0.mdx and pi05.mdx

* feat: add recompute_stats CLI operation to lerobot-edit-dataset

* fix(tests): relax quantile normalization test tolerance for 1e-6 epsilon

* chore: remove agents_memory/pr_details.md from repo

* refactor: rename delta actions to relative actions throughout

What OpenPI calls "DeltaActions" is actually UMI's "relative trajectory"
representation: each action in the chunk is an offset from the current
state, not from the previous action. This avoids error accumulation.

Renamed across all source, tests, docs, and CLI:
- DeltaActionsProcessorStep → RelativeActionsProcessorStep
- to_delta_actions → to_relative_actions
- use_delta_actions → use_relative_actions
- delta_exclude_joints → relative_exclude_joints
- compute_delta_action_stats → compute_relative_action_stats
- delta_action_processor.py → relative_action_processor.py
- test_delta_actions.py → test_relative_actions.py

Kept as-is: AbsoluteActionsProcessorStep (converts TO absolute),
registry ID "delta_actions_processor" (backward compat), and unrelated
delta references (IK pipeline, Robosuite, RA-BC metrics, gym envs).

* docs: add Action Representations guide

Dedicated page explaining absolute, relative, and delta actions with
numerical examples, joint vs EE space, and how to use kinematics
pipelines and the relative action processor. References UMI paper
(Chi et al., 2024) for the terminology.

* docs: remove redundant OpenPI naming note from action representations

* docs: remove opinionated OpenPI reference from delta actions section

* docs: replace ASCII diagram with UMI paper figure

* docs: remove OpenPI reference from action representations

* docs: use HF-hosted image instead of local asset

* docs: clarify figure attribution

* revert: restore original normalization epsilon behavior

The 1e-6 unconditional epsilon change perturbed all normalized values,
breaking backward compatibility tests. The original approach (1e-8 eps
for MEAN_STD, conditional torch.where for QUANTILES) already handles
division by zero correctly without affecting non-degenerate cases.

* fix: restore delta_action_processor.py used by phone/RL teleop

The rename commit incorrectly deleted delta_action_processor.py and
duplicated its classes into relative_action_processor.py. Restore the
original file and import from it instead.

* fix(processor): address PR #2970 review comments

- Remove shebang from relative_action_processor.py (library module, not script)
- Add device alignment in to_relative_actions/to_absolute_actions so _last_state
  on CPU doesn't cause cross-device errors when actions are on CUDA
- Rename delta_step → relative_step in AbsoluteActionsProcessorStep for naming
  consistency; update factory.py, all processor files, and tests
- Expand _reconnect_relative_absolute_steps docstring to explain why post-hoc
  rewiring is needed after deserialization
- Fix off-by-one in compute_stats.py: sample_upper_bound = total_frames - chunk_size + 1
  so last valid start index is included and total_frames == chunk_size is not rejected
- Remove redundant NOTE comment in processor_pi05.py (duplicated two lines below)
- Fix pi0_fast processor ordering: move relative_step before NormalizerProcessorStep
  so normalizer sees delta actions (matching pi0/pi05); flip postprocessor to
  unnormalize → absolute accordingly. Relative stats are now required for all pi models
- Revert use_relative_joint_actions_aloha → use_delta_joint_actions_aloha in
  configuration_smolvla.py (preserve existing public API)
- Update action_representations.mdx: add missing joint to 6-DOF example, fix
  'based on a figure', clarify pi family ordering, add RTC compatibility section

* update rtc link

* feat: compute relative action stats over full dataset with optional parallelism

Remove the 100k sample cap from compute_relative_action_stats and process
all valid chunks. Vectorize with numpy (pre-load actions/states, fancy
indexing + broadcasting) for a large speedup over the per-index HF dataset
loop. Add num_workers param for thread-based parallelism (numpy releases
the GIL). Update docs to show --push_to_hub for recompute_stats.

* style: apply ruff formatting to compute_stats.py

* testing on real robot

* style: fix ruff format and remove redundant .keys() calls

.github/workflows/fast_tests.yml

@@ -44,7 +44,7 @@ permissions:
 # Sets up the environment variables
 env:
   UV_VERSION: "0.8.0"
-  PYTHON_VERSION: "3.10"
+  PYTHON_VERSION: "3.12"
 
 # Ensures that only the latest commit for a PR or branch is built, canceling older runs.
 concurrency:
@@ -61,6 +61,7 @@ jobs:
       MUJOCO_GL: egl
       HF_HOME: /mnt/cache/.cache/huggingface
       HF_LEROBOT_HOME: /mnt/cache/.cache/huggingface/lerobot
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     steps:
       - uses: actions/checkout@v6
         with:
@@ -89,5 +90,11 @@ jobs:
       - name: Install lerobot with test extras
         run: uv sync --extra "test"
 
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          uv run hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          uv run hf auth whoami
+
       - name: Run pytest
         run: uv run pytest tests -vv --maxfail=10

.github/workflows/full_tests.yml

@@ -37,7 +37,7 @@ permissions:
 # Sets up the environment variables
 env:
   UV_VERSION: "0.8.0"
-  PYTHON_VERSION: "3.10"
+  PYTHON_VERSION: "3.12"
   DOCKER_IMAGE_NAME: huggingface/lerobot-gpu
 
 # Ensures that only the latest action is built, canceling older runs.
@@ -60,6 +60,7 @@ jobs:
       MUJOCO_GL: egl
       HF_HOME: /mnt/cache/.cache/huggingface
       HF_LEROBOT_HOME: /mnt/cache/.cache/huggingface/lerobot
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     steps:
       - uses: actions/checkout@v6
         with:
@@ -87,6 +88,12 @@ jobs:
       - name: Install lerobot with all extras
         run: uv sync --extra all # TODO(Steven): Make flash-attn optional
 
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          uv run hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          uv run hf auth whoami
+
       - name: Run pytest (all extras)
         run: uv run pytest tests -vv --maxfail=10
 
@@ -162,6 +169,7 @@ jobs:
       HF_LEROBOT_HOME: /home/user_lerobot/.cache/huggingface/lerobot
       TORCH_HOME: /home/user_lerobot/.cache/torch
       TRITON_CACHE_DIR: /home/user_lerobot/.cache/triton
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     container:
       image: ${{ needs.build-and-push-docker.outputs.image_tag }} # zizmor: ignore[unpinned-images]
       options: --gpus all --shm-size "16gb"
@@ -173,6 +181,13 @@ jobs:
         shell: bash
         working-directory: /lerobot
     steps:
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          hf auth whoami
+      - name: Fix ptxas permissions
+        run: chmod +x /lerobot/.venv/lib/python3.12/site-packages/triton/backends/nvidia/bin/ptxas
       - name: Run pytest on GPU
         run: pytest tests -vv --maxfail=10
       - name: Run end-to-end tests

.github/workflows/nightly.yml

@@ -28,7 +28,7 @@ on:
 # Sets up the environment variables
 env:
   UV_VERSION: "0.8.0"
-  PYTHON_VERSION: "3.10"
+  PYTHON_VERSION: "3.12"
   DOCKER_IMAGE_NAME_CPU: huggingface/lerobot-cpu:latest
   DOCKER_IMAGE_NAME_GPU: huggingface/lerobot-gpu:latest
 
@@ -119,6 +119,7 @@ jobs:
       HF_LEROBOT_HOME: /home/user_lerobot/.cache/huggingface/lerobot
       TORCH_HOME: /home/user_lerobot/.cache/torch
       TRITON_CACHE_DIR: /home/user_lerobot/.cache/triton
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     container:
       image: ${{ needs.build-docker-cpu-nightly.outputs.image_tag }} # zizmor: ignore[unpinned-images]
       options: --shm-size "16gb"
@@ -130,6 +131,11 @@ jobs:
         shell: bash
         working-directory: /lerobot
     steps:
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          hf auth whoami
       - name: Run pytest on CPU
         run: pytest tests -vv --maxfail=10
       - name: Run end-to-end tests
@@ -146,6 +152,7 @@ jobs:
       HF_LEROBOT_HOME: /home/user_lerobot/.cache/huggingface/lerobot
       TORCH_HOME: /home/user_lerobot/.cache/torch
       TRITON_CACHE_DIR: /home/user_lerobot/.cache/triton
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     container:
       image: ${{ needs.build-docker-gpu-nightly.outputs.image_tag }} # zizmor: ignore[unpinned-images]
       options: --gpus all --shm-size "16gb"
@@ -157,6 +164,11 @@ jobs:
         shell: bash
         working-directory: /lerobot
     steps:
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          hf auth whoami
       - name: Run pytest on GPU
         run: pytest tests -vv --maxfail=10
       - name: Run end-to-end tests
@@ -174,6 +186,7 @@ jobs:
       TORCH_HOME: /home/user_lerobot/.cache/torch
       TRITON_CACHE_DIR: /home/user_lerobot/.cache/triton
       CUDA_VISIBLE_DEVICES: "0,1,2,3"
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     container:
       image: ${{ needs.build-docker-gpu-nightly.outputs.image_tag }} # zizmor: ignore[unpinned-images]
       options: --gpus all --shm-size "16gb"
@@ -185,12 +198,15 @@ jobs:
         shell: bash
         working-directory: /lerobot
     steps:
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          hf auth whoami
       - name: Verify GPU availability
         run: |
           nvidia-smi
           python -c "import torch; print(f'PyTorch CUDA available: {torch.cuda.is_available()}'); print(f'Number of GPUs: {torch.cuda.device_count()}')"
 
       - name: Run multi-GPU training tests
-      # TODO(Steven): Investigate why motors tests are failing in multi-GPU setup
-        run: pytest tests -vv --maxfail=10 --ignore=tests/motors/
-        timeout-minutes: 10
+        run: pytest -vv tests/training/

.github/workflows/quality.yml

@@ -50,7 +50,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v6
         with:
-          python-version: '3.10'
+          python-version: '3.12'
 
       - name: Run pre-commit hooks
         uses: pre-commit/action@v3.0.1 # zizmor: ignore[unpinned-uses]

.github/workflows/release.yml

@@ -22,7 +22,7 @@ on:
 # Sets up the environment variables
 env:
   UV_VERSION: "0.8.0"
-  PYTHON_VERSION: "3.10"
+  PYTHON_VERSION: "3.12"
 
 jobs:
   # This job builds the Python package and publishes it to PyPI
@@ -45,7 +45,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v6
         with:
-          python-version: '3.10'
+          python-version: '3.12'
 
       - name: Extract Version
         id: extract_info
@@ -83,14 +83,6 @@ jobs:
             exit 1
           fi
 
-      - name: Remove Tags with Git dependencies
-        # TODO(Steven): Temporary patch to remove pi from PyPi 0.4.0 release due to its reliance on git dependencies.
-        run: |
-          echo "::info:: Checking for Git dependencies to remove from pyproject.toml..."
-          grep -E '@ git\+https|lerobot\[pi\]' pyproject.toml | sed 's/^/::warning:: Removing line: /' || true
-          sed -E -i '/@ git\+https|lerobot\[pi\]/d' pyproject.toml
-          echo "::info:: Git dependencies removed. Proceeding with build."
-
       - name: Install build dependencies
         run: python -m pip install build
 

.github/workflows/unbound_deps_tests.yml

@@ -29,7 +29,7 @@ permissions:
 # Sets up the environment variables
 env:
   UV_VERSION: "0.8.0"
-  PYTHON_VERSION: "3.10"
+  PYTHON_VERSION: "3.12"
   DOCKER_IMAGE_NAME: huggingface/lerobot-gpu:unbound
 
 # Ensures that only the latest action is built, canceling older runs.
@@ -48,6 +48,7 @@ jobs:
       MUJOCO_GL: egl
       HF_HOME: /mnt/cache/.cache/huggingface
       HF_LEROBOT_HOME: /mnt/cache/.cache/huggingface/lerobot
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     steps:
       - uses: actions/checkout@v6
         with:
@@ -79,7 +80,11 @@ jobs:
 
       - name: Install lerobot with all extras
         run: uv sync --extra all # TODO(Steven): Make flash-attn optional
-
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          uv run hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          uv run hf auth whoami
       - name: Run pytest (all extras)
         run: uv run pytest tests -vv
 
@@ -137,6 +142,7 @@ jobs:
       HF_LEROBOT_HOME: /home/user_lerobot/.cache/huggingface/lerobot
       TORCH_HOME: /home/user_lerobot/.cache/torch
       TRITON_CACHE_DIR: /home/user_lerobot/.cache/triton
+      HF_USER_TOKEN: ${{ secrets.LEROBOT_HF_USER }}
     container:
       image: ${{ needs.build-and-push-docker.outputs.image_tag }} # zizmor: ignore[unpinned-images]
       options: --gpus all --shm-size "16gb"
@@ -148,6 +154,11 @@ jobs:
         shell: bash
         working-directory: /lerobot
     steps:
+      - name: Login to Hugging Face
+        if: env.HF_USER_TOKEN != ''
+        run: |
+          hf auth login --token "$HF_USER_TOKEN" --add-to-git-credential
+          hf auth whoami
       - name: Run pytest on GPU
         run: pytest tests -vv
       - name: Run end-to-end tests

.pre-commit-config.yaml

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 default_language_version:
-    python: python3.10
+    python: python3.12
 
 exclude: "tests/artifacts/.*\\.safetensors$"
 
@@ -55,7 +55,7 @@ repos:
     rev: v3.21.0
     hooks:
     -   id: pyupgrade
-        args: [--py310-plus]
+        args: [--py312-plus]
 
   ##### Markdown Quality #####
   - repo: https://github.com/rbubley/mirrors-prettier

AI_POLICY.md

@@ -0,0 +1,25 @@
+# AI Usage Policy
+
+The LeRobot project welcomes contributions from everyone, and we have a few guidelines regarding AI usage to ensure high code quality, clear communication, and a healthy open-source ecosystem:
+
+- **Please disclose significant AI assistance.** If you used AI tools (e.g., Copilot, Claude, Cursor, ChatGPT) to generate a substantial portion of your code or text, let us know in your PR description. Transparency helps us review your changes more effectively.
+- **Own your code (The Human-in-the-Loop).** You must fully understand all the changes you are proposing. If you cannot explain what your AI-assisted code does or how it interacts with LeRobot's broader architecture, please take the time to learn and test it before submitting.
+- **Keep issues and discussions focused.** You are welcome to use AI to help draft issues or PR descriptions, but please review and edit them carefully before posting. AI can often be overly verbose; trimming the noise and getting straight to the point helps our maintainers address your needs faster.
+
+Our core maintainers also use AI tools to aid their workflows, but they do so while bringing deep contextual knowledge of the LeRobot codebase to validate the output. We ask all contributors to apply that same level of rigor.
+
+## Remember the Human Maintainers
+
+Please remember that LeRobot is maintained by a dedicated team of humans.
+
+Every discussion, issue, and pull request is read and reviewed by real people. While AI tools can generate thousands of lines of code in seconds, reviewing that code still takes human time and energy. Submitting unverified or low-effort AI output puts an unfair burden on our maintainers.
+
+Today, the quality of the AI output still heavily depends on the developer driving the tool. We ask that you respect our maintainers' time by thoroughly vetting, testing, and refining your submissions.
+
+## AI is Welcome Here
+
+LeRobot operates at the cutting edge of AI and robotics, and many of our maintainers actively embrace AI coding assistants as valuable productivity tools. We are a pro-AI project!
+
+Our reason for having an AI policy is not an anti-AI stance. Rather, it exists to ensure that AI is used to enhance human contributions, not replace them with unverified noise. It's about how the tools are used, not the tools themselves.
+
+We value the unique human insight you bring to the LeRobot community. Let AI empower your workflow, but always let your own judgment take the wheel.

CONTRIBUTING.md

@@ -2,7 +2,7 @@
 
 Everyone is welcome to contribute, and we value everybody's contribution. Code is not the only way to help the community. Answering questions, helping others, reaching out, and improving the documentation are immensely valuable.
 
-Whichever way you choose to contribute, please be mindful to respect our [code of conduct](./CODE_OF_CONDUCT.md).
+Whichever way you choose to contribute, please be mindful to respect our [code of conduct](https://github.com/huggingface/lerobot/blob/main/CODE_OF_CONDUCT.md) and our [AI policy](https://github.com/huggingface/lerobot/blob/main/AI_POLICY.md).
 
 ## Ways to Contribute
 
@@ -32,7 +32,7 @@ git remote add upstream https://github.com/huggingface/lerobot.git
 
 ### 2. Environment Installation
 
-Please follow our [Installation Guide](./docs/source/installation.mdx) for the environment setup & installation from source.
+Please follow our [Installation Guide](https://huggingface.co/docs/lerobot/installation) for the environment setup & installation from source.
 
 ## Running Tests & Quality Checks
 
@@ -75,8 +75,8 @@ pytest -sv tests/test_specific_feature.py
 
 Use the templates for required fields and examples.
 
-- **Issues:** Follow the [ticket template](./.github/ISSUE_TEMPLATE/bug-report.yml).
-- **Pull requests:** Rebase on `upstream/main`, use a descriptive branch (don't work on `main`), run `pre-commit` and tests locally, and follow the [PR template](./.github/PULL_REQUEST_TEMPLATE.md).
+- **Issues:** Follow the [ticket template](https://github.com/huggingface/lerobot/blob/main/.github/ISSUE_TEMPLATE/bug-report.yml).
+- **Pull requests:** Rebase on `upstream/main`, use a descriptive branch (don't work on `main`), run `pre-commit` and tests locally, and follow the [PR template](https://github.com/huggingface/lerobot/blob/main/.github/PULL_REQUEST_TEMPLATE.md).
 
 One member of the LeRobot team will then review your contribution.
 

MANIFEST.in

@@ -1,2 +1,3 @@
 include src/lerobot/templates/lerobot_modelcard_template.md
 include src/lerobot/datasets/card_template.md
+include src/lerobot/envs/metaworld_config.json

README.md

@@ -100,11 +100,11 @@ lerobot-train \
   --dataset.repo_id=lerobot/aloha_mobile_cabinet
 ```
 
-| Category                   | Models                                                                                                                                                                                                       |
-| -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| **Imitation Learning**     | [ACT](./docs/source/policy_act_README.md), [Diffusion](./docs/source/policy_diffusion_README.md), [VQ-BeT](./docs/source/policy_vqbet_README.md)                                                             |
-| **Reinforcement Learning** | [HIL-SERL](./docs/source/hilserl.mdx), [TDMPC](./docs/source/policy_tdmpc_README.md) & QC-FQL (coming soon)                                                                                                  |
-| **VLAs Models**            | [Pi0Fast](./docs/source/pi0fast.mdx), [Pi0.5](./docs/source/pi05.mdx), [GR00T N1.5](./docs/source/policy_groot_README.md), [SmolVLA](./docs/source/policy_smolvla_README.md), [XVLA](./docs/source/xvla.mdx) |
+| Category                   | Models                                                                                                                                                                                                                  |
+| -------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| **Imitation Learning**     | [ACT](./docs/source/policy_act_README.md), [Diffusion](./docs/source/policy_diffusion_README.md), [VQ-BeT](./docs/source/policy_vqbet_README.md), [Multitask DiT Policy](./docs/source/policy_multi_task_dit_README.md) |
+| **Reinforcement Learning** | [HIL-SERL](./docs/source/hilserl.mdx), [TDMPC](./docs/source/policy_tdmpc_README.md) & QC-FQL (coming soon)                                                                                                             |
+| **VLAs Models**            | [Pi0Fast](./docs/source/pi0fast.mdx), [Pi0.5](./docs/source/pi05.mdx), [GR00T N1.5](./docs/source/policy_groot_README.md), [SmolVLA](./docs/source/policy_smolvla_README.md), [XVLA](./docs/source/xvla.mdx)            |
 
 Similarly to the hardware, you can easily implement your own policy & leverage LeRobot's data collection, training, and visualization tools, and share your model to the HF Hub
 
@@ -135,7 +135,7 @@ Learn how to implement your own simulation environment or benchmark and distribu
 
 ## Citation
 
-If you use LeRobot in your research, please cite:
+If you use LeRobot in your project, please cite the GitHub repository to acknowledge the ongoing development and contributors:
 
 ```bibtex
 @misc{cadene2024lerobot,
@@ -146,9 +146,26 @@ If you use LeRobot in your research, please cite:
 }
 ```
 
+If you are referencing our research or the academic paper, please also cite our ICLR publication:
+
+<details>
+<summary><b>ICLR 2026 Paper</b></summary>
+
+```bibtex
+@inproceedings{cadenelerobot,
+  title={LeRobot: An Open-Source Library for End-to-End Robot Learning},
+  author={Cadene, Remi and Alibert, Simon and Capuano, Francesco and Aractingi, Michel and Zouitine, Adil and Kooijmans, Pepijn and Choghari, Jade and Russi, Martino and Pascal, Caroline and Palma, Steven and Shukor, Mustafa and Moss, Jess and Soare, Alexander and Aubakirova, Dana and Lhoest, Quentin and Gallou\'edec, Quentin and Wolf, Thomas},
+  booktitle={The Fourteenth International Conference on Learning Representations},
+  year={2026},
+  url={https://arxiv.org/abs/2602.22818}
+}
+```
+
+</details>
+
 ## Contribute
 
-We welcome contributions from everyone in the community! To get started, please read our [CONTRIBUTING.md](./CONTRIBUTING.md) guide. Whether you're adding a new feature, improving documentation, or fixing a bug, your help and feedback are invaluable. We're incredibly excited about the future of open-source robotics and can't wait to work with you on what's next—thank you for your support!
+We welcome contributions from everyone in the community! To get started, please read our [CONTRIBUTING.md](https://github.com/huggingface/lerobot/blob/main/CONTRIBUTING.md) guide. Whether you're adding a new feature, improving documentation, or fixing a bug, your help and feedback are invaluable. We're incredibly excited about the future of open-source robotics and can't wait to work with you on what's next—thank you for your support!
 
 <p align="center">
   <img alt="SO101 Video" src="./media/readme/so100_video.webp" width="640px">

benchmarks/audio/run_microphone_benchmark.py

@@ -0,0 +1,219 @@
+#!/usr/bin/env python
+
+# Copyright 2025 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 argparse
+from pathlib import Path
+
+import matplotlib.pyplot as plt
+import numpy as np
+from soundfile import read
+
+from lerobot.microphones.configs import MicrophoneConfig
+from lerobot.microphones.portaudio import PortAudioMicrophone, PortAudioMicrophoneConfig
+from lerobot.microphones.utils import (
+    async_microphones_start_recording,
+    async_microphones_stop_recording,
+    make_microphones_from_configs,
+)
+from lerobot.utils.robot_utils import (
+    precise_sleep,
+)
+
+
+def main(
+    microphones_configs: dict[str, MicrophoneConfig],
+    audio_chunks_number: int,
+    audio_chunks_duration: float,
+    repetitions: int,
+    multiprocessing: bool = False,
+):
+    recording_dir = Path("outputs/audio_benchmark")
+    recording_dir.mkdir(parents=True, exist_ok=True)
+
+    # Create microphones
+    microphones = make_microphones_from_configs(microphones_configs)
+
+    # Connect microphones
+    for microphone in microphones.values():
+        microphone.connect()
+
+    all_audio_chunks = []
+    for i in range(repetitions):
+        print(f"Repetition {i + 1}/{repetitions}...")
+
+        # Create audio chunks
+        audio_chunks = {}
+        for microphone_key in microphones:
+            audio_chunks.update({microphone_key: []})
+
+        # Start recording
+        async_microphones_start_recording(
+            microphones,
+            output_files=[
+                recording_dir / f"{microphone_key}_recording_{i}.wav" for microphone_key in microphones
+            ],
+            multiprocessing=multiprocessing,
+        )
+
+        # Record audio chunks
+        for j in range(audio_chunks_number):
+            precise_sleep(audio_chunks_duration)
+
+            for microphone_key, microphone in microphones.items():
+                audio_chunk = microphone.read()
+                print(f"{microphone_key} - repetition {i} - chunk {j} - samples {audio_chunk.shape[0]}")
+                audio_chunks[microphone_key].append(audio_chunk)
+
+        # Stop recording
+        async_microphones_stop_recording(microphones)
+
+        for microphone_key in microphones:
+            audio_chunks[microphone_key] = np.concatenate(audio_chunks[microphone_key], axis=0)
+
+        all_audio_chunks.append(audio_chunks)
+
+    # Disconnect microphones
+    for microphone in microphones.values():
+        microphone.disconnect()
+
+    # Compute statistics
+    cmap = plt.get_cmap("tab10")
+    _, ax = plt.subplots(nrows=repetitions, ncols=len(microphones))
+    chunk_length = np.zeros((repetitions, len(microphones)))
+    record_length = np.zeros((repetitions, len(microphones)))
+    for i in range(repetitions):
+        for j, (microphone_key, microphone) in enumerate(microphones.items()):
+            # Get recorded audio chunks
+            recorded_audio_chunks = all_audio_chunks[i][microphone_key]
+
+            # Load recorded file
+            recorded_data, _ = read(recording_dir / f"{microphone_key}_recording_{i}.wav")
+            if recorded_data.ndim == 1:
+                recorded_data = np.expand_dims(recorded_data, axis=1)
+
+            record_length[i, j] = recorded_data.shape[0]
+            chunk_length[i, j] = recorded_audio_chunks.shape[0]
+
+            for k, (chunk_data, record_data) in enumerate(
+                zip(recorded_audio_chunks.T, recorded_data.T, strict=False)
+            ):
+                # Plot audio chunks and recorded data
+                ax[i, j].plot(
+                    np.arange(0, len(chunk_data)) / microphone.sample_rate,
+                    chunk_data,
+                    label=f"audio chunks - channel {k}",
+                    color=cmap(2 * k),
+                )
+                ax[i, j].plot(
+                    np.arange(0, len(record_data)) / microphone.sample_rate,
+                    record_data,
+                    label=f"recorded data - channel {k}",
+                    linestyle="dashed",
+                    color=cmap(2 * k + 1),
+                )
+
+                # Plot absolute difference (errors should be located at the end of the recordings)
+                if recorded_data.shape[0] - recorded_audio_chunks.shape[0] > 0:
+                    chunk_data = np.append(
+                        chunk_data, np.zeros(int(recorded_data.shape[0] - recorded_audio_chunks.shape[0]))
+                    )
+                else:
+                    record_data = np.append(
+                        record_data, np.zeros(int(-recorded_data.shape[0] + recorded_audio_chunks.shape[0]))
+                    )
+                ax[i, j].plot(
+                    np.arange(0, len(record_data)) / microphone.sample_rate,
+                    np.abs(chunk_data - record_data),
+                    label=f"differences - channel {k}",
+                    color="red",
+                    linestyle="dotted",
+                )
+                ax[i, j].set_title(f"{microphone_key} - repetition {i}")
+            ax[i, j].legend()
+
+    plt.show()
+
+    # Print statistics
+    differences = record_length - chunk_length
+    for i, (microphone_key, microphone) in enumerate(microphones.items()):
+        print(
+            f"Average recorded duration for {microphone_key} : {np.mean(record_length[:, i]) / microphone.sample_rate:.3f} seconds"
+        )
+        print(
+            f"Average chunk duration for {microphone_key} : {np.mean(chunk_length[:, i]) / microphone.sample_rate:.3f} seconds"
+        )
+        print(f"Average difference for {microphone_key} : {np.mean(differences[:, i]):.3f} samples")
+        print(
+            f"Average difference for {microphone_key} : {np.mean(differences[:, i]) / microphone.sample_rate:.3f} seconds"
+        )
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--microphones_indices",
+        type=int,
+        nargs="+",
+        default=[microphone["index"] for microphone in PortAudioMicrophone.find_microphones()],
+    )
+    parser.add_argument(
+        "--microphones_sample_rate",
+        type=float,
+        nargs="+",
+        default=[None] * len(PortAudioMicrophone.find_microphones()),
+    )
+    parser.add_argument(
+        "--microphones_channels",
+        type=int,
+        nargs="+",
+        default=[None] * len(PortAudioMicrophone.find_microphones()),
+    )
+    parser.add_argument("--audio_chunks_number", type=int, default=2)
+    parser.add_argument(
+        "--audio_chunks_duration",
+        type=float,
+        default=1.0,
+    )
+    parser.add_argument(
+        "--repetitions",
+        type=int,
+        default=2,
+    )
+    parser.add_argument(
+        "--multiprocessing",
+        action="store_true",
+    )
+
+    args = vars(parser.parse_args())
+
+    args["microphones_configs"] = {}
+    for index, sample_rate, channels in zip(
+        args["microphones_indices"],
+        args["microphones_sample_rate"],
+        args["microphones_channels"],
+        strict=False,
+    ):
+        microphone_config = PortAudioMicrophoneConfig(
+            microphone_index=index,
+            sample_rate=sample_rate,
+            channels=channels,
+        )
+        args["microphones_configs"].update({f"microphone_{index}": microphone_config})
+    args.pop("microphones_indices")
+    args.pop("microphones_sample_rate")
+    args.pop("microphones_channels")
+
+    main(**args)

benchmarks/audio/run_tactile_benchmark.py

@@ -0,0 +1,137 @@
+#!/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 argparse
+from pathlib import Path
+
+import numpy as np
+import soundfile as sf
+
+from lerobot.microphones.configs import MicrophoneConfig
+from lerobot.microphones.touchlab import TouchLabSensorConfig
+from lerobot.microphones.utils import (
+    async_microphones_start_recording,
+    async_microphones_stop_recording,
+    make_microphones_from_configs,
+)
+from lerobot.utils.robot_utils import (
+    precise_sleep,
+)
+
+
+def main(
+    sensors_configs: dict[str, MicrophoneConfig],
+    multiprocessing: bool = False,
+):
+    recording_dir = Path("outputs/tactile_benchmark")
+    recording_dir.mkdir(parents=True, exist_ok=True)
+
+    # Create microphones
+    sensors = make_microphones_from_configs(sensors_configs)
+
+    # Connect microphones
+    for sensor in sensors.values():
+        sensor.connect()
+
+    # Create audio chunks
+    data_chunks = {}
+    for sensor_key in sensors:
+        data_chunks.update({sensor_key: []})
+
+    # Start recording
+    async_microphones_start_recording(
+        sensors,
+        output_files=[recording_dir / f"{sensor_key}_recording.wav" for sensor_key in sensors],
+        multiprocessing=multiprocessing,
+    )
+
+    # Record audio chunks
+    precise_sleep(10.0)
+
+    for sensor_key, sensor in sensors.items():
+        data_chunk = sensor.read()
+        print(f"{sensor_key} - samples {data_chunk.shape[0]}")
+        data_chunks[sensor_key].append(data_chunk)
+
+    # Stop recording
+    async_microphones_stop_recording(sensors)
+
+    for sensor_key in sensors:
+        data_chunks[sensor_key] = np.concatenate(data_chunks[sensor_key], axis=0)
+
+    # Disconnect microphones
+    for sensor in sensors.values():
+        sensor.disconnect()
+
+    for sensor_key in sensors:
+        data, sample_rate = sf.read(recording_dir / f"{sensor_key}_recording.wav")
+        print(f"{sensor_key} - samples {data.shape[0]}")
+        print(f"{sensor_key} - sample rate {sample_rate}")
+        print(f"{sensor_key} - data {data}")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--sensors_ports",
+        type=str,
+        nargs="+",
+    )
+    parser.add_argument(
+        "--sensors_baud_rate",
+        type=int,
+        nargs="+",
+    )
+    parser.add_argument(
+        "--sensors_sample_rate",
+        type=int,
+        nargs="+",
+    )
+    parser.add_argument(
+        "--sensors_channels",
+        type=int,
+        nargs="+",
+    )
+    parser.add_argument(
+        "--multiprocessing",
+        action="store_true",
+    )
+
+    args = vars(parser.parse_args())
+
+    args["sensors_configs"] = {}
+    for port, baud_rate, sample_rate, channels in zip(
+        args["sensors_ports"],
+        args["sensors_baud_rate"],
+        args["sensors_sample_rate"],
+        args["sensors_channels"],
+        strict=False,
+    ):
+        if isinstance(channels, int):
+            channels = [channels]
+        sensor_config = TouchLabSensorConfig(
+            sensor_port=port,
+            baud_rate=baud_rate,
+            sample_rate=sample_rate,
+            channels=channels,
+        )
+        args["sensors_configs"].update({f"sensor_{port}": sensor_config})
+    args.pop("sensors_ports")
+    args.pop("sensors_baud_rate")
+    args.pop("sensors_sample_rate")
+    args.pop("sensors_channels")
+
+    main(**args)

benchmarks/tokens/README.md

@@ -1,134 +0,0 @@
-# Action tokenizer benchmark
-
-## Questions
-
-What is the trade-off between:
-
-- **Compression**: how many tokens are needed to represent an action chunk (e.g. horizon × action_dim floats)?
-- **Reconstruction quality**: how well does encode-then-decode preserve the original actions?
-- **Speed**: how long does encoding and decoding take per chunk?
-
-How to choose an action tokenizer?
-
-- Which tokenizer architecture (e.g. dct + BPE, DCT + BPE)?
-- Which **action horizon** and **encoded dimensions** to use?
-- Which **normalization** (QUANTILES, MEAN_STD, MIN_MAX) and **delta transform** (relative vs absolute actions)?
-- How do reconstruction error and compression ratio vary across datasets and tokenizer settings?
-
-This benchmark loads action chunks from a LeRobot dataset using the same pipeline as `lerobot-train-tokenizer`, runs a trained action tokenizer in encode/decode mode, and reports reconstruction error, compression stats, and timing. Results are saved as JSON under `outputs/` for comparison and analysis.
-
-## Variables
-
-**Dataset & chunking**
-
-- **repo_id**: LeRobot dataset (e.g. `lerobot/pusht`). Action statistics and normalization are taken from the dataset metadata when available.
-- **action_horizon**: Number of future steps per action chunk (must match the tokenizer’s training).
-- **encoded_dims**: Dimension ranges to encode (e.g. `0:6` or `0:6,7:14`). Must match the tokenizer.
-- **max_episodes**: Cap on episodes to load (default: all).
-- **sample_fraction**: Fraction of chunks to sample per episode (default `0.2`) to keep runtime manageable.
-
-**Transform & normalization**
-
-- **normalization_mode**: `IDENTITY`, `MEAN_STD`, `MIN_MAX`, `QUANTILES`, `QUANTILE10`. Should match the tokenizer’s training.
-- **delta_dims**: Comma-separated dimension indices for delta (relative) transform.
-- **use_delta_transform**: Whether to convert actions to relative to current state for those dimensions.
-- **state_key**: Dataset key for state (e.g. `observation.state`) used when applying delta transform.
-
-**Tokenizer & evaluation**
-
-- **action_tokenizer_path**: Path or HuggingFace repo id of the trained tokenizer (e.g. `outputs/wavetoken`).
-- **max_chunks_for_reconstruction**: Max number of chunks to use for reconstruction and timing (default `500`) to limit runtime.
-
-### Main parameters
-
-| parameter                        | default                      | description                                      |
-| -------------------------------- | ---------------------------- | ------------------------------------------------ |
-| **action_tokenizer_path**        | (required)                   | Path or Hub id of the trained action tokenizer.  |
-| **repo_id**                      | (required)                   | LeRobot dataset repo id.                         |
-| **action_horizon**               | `10`                         | Future steps per chunk.                          |
-| **encoded_dims**                 | `0:6`                        | Dimension ranges to encode (e.g. `0:6,7:14`).   |
-| **normalization_mode**           | `QUANTILES`                  | Normalization mode for actions.                  |
-| **max_episodes**                 | all                          | Max episodes to load.                            |
-| **sample_fraction**              | `0.2`                        | Fraction of chunks sampled per episode.          |
-| **max_chunks_for_reconstruction**| `500`                        | Chunks used for reconstruction and timing.       |
-| **output_dir**                   | `outputs/action_tokenizer_benchmark` | Directory for results JSON.              |
-
-## Metrics
-
-**Reconstruction (lower is better)**
-
-- **reconstruction_mae**: Mean absolute error between original and decoded action chunks.
-- **reconstruction_mse**: Mean squared error.
-- **reconstruction_rmse**: Root mean squared error.
-- **reconstruction_max_abs_error**: Maximum absolute error over all dimensions and samples.
-- **per_dimension_mae**: MAE per action dimension (list of length `action_dim`).
-
-**Compression**
-
-- **compression_ratio**: Ratio (action_horizon × action_dim) / mean number of tokens. Higher means more compression.
-- **mean_token_length**, **std_token_length**: Mean and standard deviation of token count per chunk.
-- **min_token_length**, **max_token_length**: Min and max token count.
-- **p50_token_length**, **p99_token_length**: 50th and 99th percentile token counts.
-
-**Timing (seconds per chunk)**
-
-- **mean_encode_time_sec**: Mean time to encode one chunk.
-- **mean_decode_time_sec**: Mean time to decode one chunk.
-
-The JSON output also includes **num_chunks_evaluated** and **total_chunks_available** for context.
-
-## How the benchmark works
-
-1. **Load dataset**: LeRobot dataset is loaded for the given `repo_id` and `root`.
-2. **Build action chunks**: For each episode (up to `max_episodes`), action chunks are built with the same logic as `lerobot-train-tokenizer`: sliding window of length `action_horizon`, optional delta transform, and per-episode sampling with `sample_fraction`.
-3. **Extract and normalize**: Only `encoded_dims` are kept. Normalization is applied using the dataset’s action stats when available, according to `normalization_mode`.
-4. **Encode / decode**: A random sample of chunks (size `max_chunks_for_reconstruction`) is encoded and then decoded with the tokenizer. Encode and decode times are recorded per chunk.
-5. **Compute metrics**: Reconstruction metrics are computed between original and decoded chunks; compression and timing stats are aggregated.
-6. **Save results**: A JSON file is written to `output_dir` with name `{timestamp}_{repo_id}_action_tokenizer_results.json`, containing the full config and all metrics.
-
-The pipeline (chunking, dimensions, normalization, delta) must match how the tokenizer was trained; otherwise reconstruction error can be large or the tokenizer may raise.
-
-## Caveats
-
-- The tokenizer’s **action_horizon** and **action_dim** (and optionally DCT settings) are fixed at training time. The benchmark infers dimensions from the dataset and encoded dims; the tokenizer path must correspond to a model trained with the same horizon and encoded dimensions.
-- Reconstruction is evaluated in **normalized space** (the same space the tokenizer sees). For interpretation in raw action space, you would need to invert normalization outside this script.
-- Only one tokenizer and one dataset are evaluated per run. To compare tokenizers or datasets, run the script multiple times and compare the saved JSON files.
-
-## Example
-
-Quick run with a local tokenizer and a small number of episodes:
-
-```bash
-python benchmarks/tokens/run_action_tokenizer_benchmark.py \
-    --action-tokenizer-path=outputs/wavetoken \
-    --repo-id=lerobot/pusht \
-    --action-horizon=10 \
-    --max-episodes=50 \
-    --output-dir=outputs/action_tokenizer_benchmark
-```
-
-With delta transform and custom encoded dimensions:
-
-```bash
-python benchmarks/tokens/run_action_tokenizer_benchmark.py \
-    --action-tokenizer-path=outputs/wavetoken \
-    --repo-id=lerobot/pusht \
-    --action-horizon=10 \
-    --encoded-dims=0:6,7:14 \
-    --delta-dims=0,1,2,3,4,5 \
-    --use-delta-transform \
-    --normalization-mode=QUANTILES \
-    --max-chunks-for-reconstruction=500 \
-    --output-dir=outputs/action_tokenizer_benchmark
-```
-
-Results are written to e.g. `outputs/action_tokenizer_benchmark/2026-02-12_14-30-00_lerobot_pusht_action_tokenizer_results.json`.
-
-## Results
-
-Results are stored as JSON in the directory given by `--output-dir` (default: `outputs/action_tokenizer_benchmark`). Each file contains:
-
-- **config**: All script arguments (tokenizer path, repo_id, action_horizon, encoded_dims, normalization_mode, etc.) for reproducibility.
-- **metrics**: All reconstruction, compression, and timing metrics described above.
-
-To compare runs, load and diff or aggregate these JSON files with your own scripts or notebooks.

benchmarks/tokens/run_action_tokenizer_benchmark.py

@@ -1,442 +0,0 @@
-#!/usr/bin/env python
-# Copyright 2026 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.
-"""Benchmark action tokenization: reconstruction error, compression ratio, and timing.
-
-Loads action chunks from a LeRobot dataset, encodes/decodes them with a trained action
-tokenizer, and reports:
-- Reconstruction: MAE, MSE, RMSE, max absolute error, per-dimension MAE
-- Jerk: mean absolute jerk (original and reconstructed), jerk reconstruction MAE
-- Compression: ratio (input size / mean tokens), token length stats
-- Timing: mean encode/decode time per chunk
-
-Results are saved to outputs/action_tokenizer_benchmark/<timestamp>_results.json.
-
-Example:
-
-```bash
-python benchmarks/tokens/run_action_tokenizer_benchmark.py \
-    --action-tokenizer-path=outputs/wavetoken \
-    --repo-id=lerobot/pusht \
-    --action-horizon=10 \
-    --max-episodes=50 \
-    --output-dir=outputs/action_tokenizer_benchmark
-```
-"""
-
-import argparse
-import json
-import time
-from pathlib import Path
-
-import numpy as np
-
-from lerobot.configs.types import NormalizationMode
-from lerobot.datasets.lerobot_dataset import LeRobotDataset
-from lerobot.utils.constants import ACTION, OBS_STATE
-
-# Optional: use same helpers as train script if we want to avoid duplication
-from lerobot.scripts.lerobot_train_tokenizer import (
-    apply_normalization,
-    process_episode,
-)
-
-
-def load_action_chunks(
-    repo_id: str,
-    root: str | None,
-    action_horizon: int,
-    max_episodes: int | None,
-    sample_fraction: float,
-    encoded_dims: str,
-    delta_dims: str | None,
-    use_delta_transform: bool,
-    state_key: str,
-    normalization_mode: NormalizationMode,
-):
-    """Load and normalize action chunks from a LeRobot dataset (same pipeline as training)."""
-    dataset = LeRobotDataset(repo_id=repo_id, root=root)
-    num_episodes = dataset.num_episodes
-    if max_episodes is not None:
-        num_episodes = min(max_episodes, num_episodes)
-
-    # Parse encoded dims
-    encoded_dim_ranges = []
-    for range_str in encoded_dims.split(","):
-        start, end = map(int, range_str.strip().split(":"))
-        encoded_dim_ranges.append((start, end))
-    total_encoded_dims = sum(end - start for start, end in encoded_dim_ranges)
-
-    delta_dim_list = None
-    if delta_dims is not None and delta_dims.strip():
-        delta_dim_list = [int(d.strip()) for d in delta_dims.split(",")]
-
-    all_chunks = []
-    for ep_idx in range(num_episodes):
-        chunks = process_episode(
-            (
-                dataset,
-                ep_idx,
-                action_horizon,
-                delta_dim_list,
-                sample_fraction,
-                state_key,
-                use_delta_transform,
-            )
-        )
-        if chunks is not None:
-            all_chunks.append(chunks)
-
-    if not all_chunks:
-        raise ValueError("No action chunks collected. Check action_horizon and dataset.")
-
-    all_chunks = np.concatenate(all_chunks, axis=0)
-
-    # Extract encoded dimensions only
-    encoded_chunks = []
-    for start, end in encoded_dim_ranges:
-        encoded_chunks.append(all_chunks[:, :, start:end])
-    encoded_chunks = np.concatenate(encoded_chunks, axis=-1)
-
-    # Normalize
-    norm_stats = dataset.meta.stats
-    if norm_stats is not None and ACTION in norm_stats:
-        action_stats = norm_stats[ACTION]
-        encoded_dim_indices = []
-        for start, end in encoded_dim_ranges:
-            encoded_dim_indices.extend(range(start, end))
-        encoded_dim_indices = np.array(encoded_dim_indices)
-        encoded_stats = {}
-        for stat_name, stat_values in action_stats.items():
-            if isinstance(stat_values, (list, np.ndarray)):
-                stat_array = np.array(stat_values)
-                if len(stat_array) > max(encoded_dim_indices):
-                    encoded_stats[stat_name] = stat_array[encoded_dim_indices]
-        if encoded_stats:
-            try:
-                encoded_chunks = apply_normalization(
-                    encoded_chunks, encoded_stats, normalization_mode, eps=1e-8
-                )
-            except ValueError:
-                pass
-
-    return encoded_chunks, total_encoded_dims, action_horizon, dataset.repo_id
-
-
-def compute_reconstruction_metrics(original: np.ndarray, reconstructed: np.ndarray):
-    """Compute reconstruction error metrics (original and reconstructed same shape [N, T, D])."""
-    diff = reconstructed - original
-    mae = float(np.mean(np.abs(diff)))
-    mse = float(np.mean(diff**2))
-    rmse = float(np.sqrt(mse))
-    max_abs_err = float(np.max(np.abs(diff)))
-
-    # Per-dimension MAE (over N and T)
-    per_dim_mae = np.mean(np.abs(diff), axis=(0, 1))
-    per_dim_mae = per_dim_mae.tolist()
-
-    return {
-        "reconstruction_mae": mae,
-        "reconstruction_mse": mse,
-        "reconstruction_rmse": rmse,
-        "reconstruction_max_abs_error": max_abs_err,
-        "per_dimension_mae": per_dim_mae,
-    }
-
-
-def compute_jerk_metrics(original: np.ndarray, reconstructed: np.ndarray) -> dict:
-    """Compute jerk (3rd derivative of action w.r.t. time) metrics.
-
-    Args:
-        original: Action chunks [N, T, D].
-        reconstructed: Reconstructed action chunks [N, T, D].
-
-    Returns:
-        Dict with mean absolute jerk for original, reconstructed, and jerk reconstruction MAE.
-    """
-    # Jerk = 3rd discrete difference along time axis; need T >= 4
-    if original.shape[1] < 4:
-        return {}
-    jerk_orig = np.diff(original, n=3, axis=1)  # (N, T-3, D)
-    jerk_recon = np.diff(reconstructed, n=3, axis=1)
-    mae_jerk_orig = float(np.mean(np.abs(jerk_orig)))
-    mae_jerk_recon = float(np.mean(np.abs(jerk_recon)))
-    jerk_reconstruction_mae = float(np.mean(np.abs(jerk_recon - jerk_orig)))
-    return {
-        "jerk_mae_original": mae_jerk_orig,
-        "jerk_mae_reconstructed": mae_jerk_recon,
-        "jerk_reconstruction_mae": jerk_reconstruction_mae,
-    }
-
-
-def run_benchmark(
-    action_chunks: np.ndarray,
-    action_horizon: int,
-    action_dim: int,
-    tokenizer_path: str,
-    max_chunks_for_reconstruction: int | None = 500,
-):
-    """Encode/decode action chunks and compute metrics."""
-    from transformers import AutoProcessor
-
-    processor = AutoProcessor.from_pretrained(tokenizer_path, trust_remote_code=True)
-
-    n_chunks = len(action_chunks)
-    sample_size = n_chunks
-    if max_chunks_for_reconstruction is not None:
-        sample_size = min(max_chunks_for_reconstruction, n_chunks)
-    rng = np.random.RandomState(42)
-    indices = rng.choice(n_chunks, size=sample_size, replace=False)
-    sample_chunks = action_chunks[indices]
-
-    # Encode
-    token_lengths = []
-    encode_times = []
-    all_tokens = []
-    for i in range(len(sample_chunks)):
-        chunk = sample_chunks[i : i + 1]
-        t0 = time.perf_counter()
-        tokens = processor(chunk)[0]
-        encode_times.append(time.perf_counter() - t0)
-        if isinstance(tokens, list):
-            token_lengths.append(len(tokens))
-            all_tokens.append(tokens)
-        else:
-            n = tokens.shape[0] if hasattr(tokens, "shape") else len(tokens)
-            token_lengths.append(n)
-            all_tokens.append(tokens.tolist() if hasattr(tokens, "tolist") else list(tokens))
-
-    # Decode (processor keeps time_horizon/action_dim from encode)
-    decoded_list = []
-    decode_times = []
-    for i, tok_list in enumerate(all_tokens):
-        t0 = time.perf_counter()
-        recon = processor.decode(
-            [tok_list],
-            time_horizon=action_horizon,
-            action_dim=action_dim,
-        )
-        decode_times.append(time.perf_counter() - t0)
-        decoded_list.append(recon)
-    decoded = np.concatenate(decoded_list, axis=0)
-
-    # Reconstruction metrics
-    metrics = compute_reconstruction_metrics(sample_chunks, decoded)
-
-    # Jerk metrics (3rd derivative along time)
-    jerk_metrics = compute_jerk_metrics(sample_chunks, decoded)
-    metrics.update(jerk_metrics)
-
-    # Compression
-    token_lengths = np.array(token_lengths)
-    input_size = action_horizon * action_dim
-    compression_ratio = input_size / float(np.mean(token_lengths))
-    metrics["compression_ratio"] = compression_ratio
-    metrics["mean_token_length"] = float(np.mean(token_lengths))
-    metrics["std_token_length"] = float(np.std(token_lengths))
-    metrics["min_token_length"] = int(np.min(token_lengths))
-    metrics["max_token_length"] = int(np.max(token_lengths))
-    metrics["p50_token_length"] = float(np.percentile(token_lengths, 50))
-    metrics["p99_token_length"] = float(np.percentile(token_lengths, 99))
-
-    # Timing (seconds per chunk)
-    metrics["mean_encode_time_sec"] = float(np.mean(encode_times))
-    metrics["mean_decode_time_sec"] = float(np.mean(decode_times))
-    metrics["num_chunks_evaluated"] = sample_size
-    metrics["total_chunks_available"] = n_chunks
-
-    return metrics
-
-
-def main(
-    action_tokenizer_path: str,
-    repo_id: str,
-    root: str | None = None,
-    action_horizon: int = 10,
-    max_episodes: int | None = 100,
-    sample_fraction: float = 0.2,
-    encoded_dims: str = "0:6",
-    delta_dims: str | None = None,
-    use_delta_transform: bool = False,
-    state_key: str = OBS_STATE,
-    normalization_mode: str = "QUANTILES",
-    max_chunks_for_reconstruction: int | None = 500,
-    output_dir: str | None = None,
-):
-    if output_dir is None:
-        output_dir = "outputs/action_tokenizer_benchmark"
-    output_path = Path(output_dir)
-    output_path.mkdir(parents=True, exist_ok=True)
-
-    try:
-        norm_mode = NormalizationMode(normalization_mode)
-    except ValueError:
-        norm_mode = NormalizationMode.QUANTILES
-
-    print("Loading action chunks...")
-    encoded_chunks, action_dim, horizon, _ = load_action_chunks(
-        repo_id=repo_id,
-        root=root,
-        action_horizon=action_horizon,
-        max_episodes=max_episodes,
-        sample_fraction=sample_fraction,
-        encoded_dims=encoded_dims,
-        delta_dims=delta_dims,
-        use_delta_transform=use_delta_transform,
-        state_key=state_key,
-        normalization_mode=norm_mode,
-    )
-    print(f"Loaded {len(encoded_chunks)} chunks, shape {encoded_chunks.shape} (H={horizon}, D={action_dim})")
-
-    print("Running tokenizer benchmark...")
-    metrics = run_benchmark(
-        action_chunks=encoded_chunks,
-        action_horizon=horizon,
-        action_dim=action_dim,
-        tokenizer_path=action_tokenizer_path,
-        max_chunks_for_reconstruction=max_chunks_for_reconstruction,
-    )
-
-    # Attach config for reproducibility
-    results = {
-        "config": {
-            "action_tokenizer_path": action_tokenizer_path,
-            "repo_id": repo_id,
-            "action_horizon": action_horizon,
-            "max_episodes": max_episodes,
-            "sample_fraction": sample_fraction,
-            "encoded_dims": encoded_dims,
-            "delta_dims": delta_dims,
-            "use_delta_transform": use_delta_transform,
-            "state_key": state_key,
-            "normalization_mode": normalization_mode,
-        },
-        "metrics": metrics,
-    }
-
-    timestamp = time.strftime("%Y-%m-%d_%H-%M-%S")
-    safe_repo = repo_id.replace("/", "_")
-    out_file = output_path / f"{timestamp}_{safe_repo}_action_tokenizer_results.json"
-    with open(out_file, "w") as f:
-        json.dump(results, f, indent=2)
-
-    print(f"Results saved to {out_file}")
-    print("Metrics:")
-    for k, v in metrics.items():
-        if isinstance(v, list):
-            print(f"  {k}: (length {len(v)})")
-        else:
-            print(f"  {k}: {v}")
-
-    return results
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description="Benchmark action tokenization (reconstruction error, compression, timing)."
-    )
-    parser.add_argument(
-        "--action-tokenizer-path",
-        type=str,
-        required=True,
-        help="Path or HuggingFace repo id of the trained action tokenizer (e.g. outputs/wavetoken).",
-    )
-    parser.add_argument(
-        "--repo-id",
-        type=str,
-        required=True,
-        help="LeRobot dataset repo id (e.g. lerobot/pusht).",
-    )
-    parser.add_argument(
-        "--root",
-        type=str,
-        default=None,
-        help="Root directory for LeRobot datasets.",
-    )
-    parser.add_argument(
-        "--action-horizon",
-        type=int,
-        default=10,
-        help="Number of future steps per action chunk.",
-    )
-    parser.add_argument(
-        "--max-episodes",
-        type=int,
-        default=None,
-        help="Max episodes to use (default: all).",
-    )
-    parser.add_argument(
-        "--sample-fraction",
-        type=float,
-        default=0.2,
-        help="Fraction of chunks to sample per episode.",
-    )
-    parser.add_argument(
-        "--encoded-dims",
-        type=str,
-        default="0:6",
-        help="Dimension ranges to encode (e.g. 0:6,7:14).",
-    )
-    parser.add_argument(
-        "--delta-dims",
-        type=str,
-        default=None,
-        help="Comma-separated dimensions for delta transform.",
-    )
-    parser.add_argument(
-        "--use-delta-transform",
-        action="store_true",
-        help="Apply delta (relative) transform to specified dimensions.",
-    )
-    parser.add_argument(
-        "--state-key",
-        type=str,
-        default=OBS_STATE,
-        help="Dataset key for state (for delta transform).",
-    )
-    parser.add_argument(
-        "--normalization-mode",
-        type=str,
-        default="QUANTILES",
-        choices=[m.value for m in NormalizationMode],
-        help="Normalization mode for actions.",
-    )
-    parser.add_argument(
-        "--max-chunks-for-reconstruction",
-        type=int,
-        default=500,
-        help="Max chunks to use for reconstruction metrics (default: 500).",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=str,
-        default="outputs/action_tokenizer_benchmark",
-        help="Directory to save results JSON (default: outputs/action_tokenizer_benchmark).",
-    )
-    args = parser.parse_args()
-    main(
-        action_tokenizer_path=args.action_tokenizer_path,
-        repo_id=args.repo_id,
-        root=args.root,
-        action_horizon=args.action_horizon,
-        max_episodes=args.max_episodes,
-        sample_fraction=args.sample_fraction,
-        encoded_dims=args.encoded_dims,
-        delta_dims=args.delta_dims,
-        use_delta_transform=args.use_delta_transform,
-        state_key=args.state_key,
-        normalization_mode=args.normalization_mode,
-        max_chunks_for_reconstruction=args.max_chunks_for_reconstruction,
-        output_dir=args.output_dir,
-    )

benchmarks/video/README.md

@@ -28,9 +28,9 @@ We don't expect the same optimal settings for a dataset of images from a simulat
 For these reasons, we run this benchmark on four representative datasets:
 
 - `lerobot/pusht_image`: (96 x 96 pixels) simulation with simple geometric shapes, fixed camera.
-- `aliberts/aloha_mobile_shrimp_image`: (480 x 640 pixels) real-world indoor, moving camera.
-- `aliberts/paris_street`: (720 x 1280 pixels) real-world outdoor, moving camera.
-- `aliberts/kitchen`: (1080 x 1920 pixels) real-world indoor, fixed camera.
+- `lerobot/aloha_mobile_shrimp_image`: (480 x 640 pixels) real-world indoor, moving camera.
+- `lerobot/paris_street`: (720 x 1280 pixels) real-world outdoor, moving camera.
+- `lerobot/kitchen`: (1080 x 1920 pixels) real-world indoor, fixed camera.
 
 Note: The datasets used for this benchmark need to be image datasets, not video datasets.
 
@@ -179,7 +179,7 @@ python benchmark/video/run_video_benchmark.py \
     --output-dir outputs/video_benchmark \
     --repo-ids \
         lerobot/pusht_image \
-        aliberts/aloha_mobile_shrimp_image \
+        lerobot/aloha_mobile_shrimp_image \
     --vcodec libx264 libx265 \
     --pix-fmt yuv444p yuv420p \
     --g 2 20 None \
@@ -203,9 +203,9 @@ python benchmark/video/run_video_benchmark.py \
     --output-dir outputs/video_benchmark \
     --repo-ids \
         lerobot/pusht_image \
-        aliberts/aloha_mobile_shrimp_image \
-        aliberts/paris_street \
-        aliberts/kitchen \
+        lerobot/aloha_mobile_shrimp_image \
+        lerobot/paris_street \
+        lerobot/kitchen \
     --vcodec libx264 libx265 \
     --pix-fmt yuv444p yuv420p \
     --g 1 2 3 4 5 6 10 15 20 40 None \
@@ -221,9 +221,9 @@ python benchmark/video/run_video_benchmark.py \
     --output-dir outputs/video_benchmark \
     --repo-ids \
         lerobot/pusht_image \
-        aliberts/aloha_mobile_shrimp_image \
-        aliberts/paris_street \
-        aliberts/kitchen \
+        lerobot/aloha_mobile_shrimp_image \
+        lerobot/paris_street \
+        lerobot/kitchen \
     --vcodec libsvtav1 \
     --pix-fmt yuv420p \
     --g 1 2 3 4 5 6 10 15 20 40 None \
@@ -252,37 +252,37 @@ Since we're using av1 encoding, we're choosing the `pyav` decoder as `video_read
 
 These tables show the results for `g=2` and `crf=30`, using `timestamps-modes=6_frames` and `backend=pyav`
 
-| video_images_size_ratio            | vcodec     | pix_fmt |           |           |           |
-| ---------------------------------- | ---------- | ------- | --------- | --------- | --------- |
-|                                    | libx264    |         | libx265   |           | libsvtav1 |
-| repo_id                            | yuv420p    | yuv444p | yuv420p   | yuv444p   | yuv420p   |
-| lerobot/pusht_image                | **16.97%** | 17.58%  | 18.57%    | 18.86%    | 22.06%    |
-| aliberts/aloha_mobile_shrimp_image | 2.14%      | 2.11%   | 1.38%     | **1.37%** | 5.59%     |
-| aliberts/paris_street              | 2.12%      | 2.13%   | **1.54%** | **1.54%** | 4.43%     |
-| aliberts/kitchen                   | 1.40%      | 1.39%   | **1.00%** | **1.00%** | 2.52%     |
+| video_images_size_ratio           | vcodec     | pix_fmt |           |           |           |
+| --------------------------------- | ---------- | ------- | --------- | --------- | --------- |
+|                                   | libx264    |         | libx265   |           | libsvtav1 |
+| repo_id                           | yuv420p    | yuv444p | yuv420p   | yuv444p   | yuv420p   |
+| lerobot/pusht_image               | **16.97%** | 17.58%  | 18.57%    | 18.86%    | 22.06%    |
+| lerobot/aloha_mobile_shrimp_image | 2.14%      | 2.11%   | 1.38%     | **1.37%** | 5.59%     |
+| lerobot/paris_street              | 2.12%      | 2.13%   | **1.54%** | **1.54%** | 4.43%     |
+| lerobot/kitchen                   | 1.40%      | 1.39%   | **1.00%** | **1.00%** | 2.52%     |
 
-| video_images_load_time_ratio       | vcodec  | pix_fmt |          |         |           |
-| ---------------------------------- | ------- | ------- | -------- | ------- | --------- |
-|                                    | libx264 |         | libx265  |         | libsvtav1 |
-| repo_id                            | yuv420p | yuv444p | yuv420p  | yuv444p | yuv420p   |
-| lerobot/pusht_image                | 6.45    | 5.19    | **1.90** | 2.12    | 2.47      |
-| aliberts/aloha_mobile_shrimp_image | 11.80   | 7.92    | 0.71     | 0.85    | **0.48**  |
-| aliberts/paris_street              | 2.21    | 2.05    | 0.36     | 0.49    | **0.30**  |
-| aliberts/kitchen                   | 1.46    | 1.46    | 0.28     | 0.51    | **0.26**  |
+| video_images_load_time_ratio      | vcodec  | pix_fmt |          |         |           |
+| --------------------------------- | ------- | ------- | -------- | ------- | --------- |
+|                                   | libx264 |         | libx265  |         | libsvtav1 |
+| repo_id                           | yuv420p | yuv444p | yuv420p  | yuv444p | yuv420p   |
+| lerobot/pusht_image               | 6.45    | 5.19    | **1.90** | 2.12    | 2.47      |
+| lerobot/aloha_mobile_shrimp_image | 11.80   | 7.92    | 0.71     | 0.85    | **0.48**  |
+| lerobot/paris_street              | 2.21    | 2.05    | 0.36     | 0.49    | **0.30**  |
+| lerobot/kitchen                   | 1.46    | 1.46    | 0.28     | 0.51    | **0.26**  |
 
-|                                    |          | vcodec   | pix_fmt      |          |           |              |
-| ---------------------------------- | -------- | -------- | ------------ | -------- | --------- | ------------ |
-|                                    |          | libx264  |              | libx265  |           | libsvtav1    |
-| repo_id                            | metric   | yuv420p  | yuv444p      | yuv420p  | yuv444p   | yuv420p      |
-| lerobot/pusht_image                | avg_mse  | 2.90E-04 | **2.03E-04** | 3.13E-04 | 2.29E-04  | 2.19E-04     |
-|                                    | avg_psnr | 35.44    | 37.07        | 35.49    | **37.30** | 37.20        |
-|                                    | avg_ssim | 98.28%   | **98.85%**   | 98.31%   | 98.84%    | 98.72%       |
-| aliberts/aloha_mobile_shrimp_image | avg_mse  | 2.76E-04 | 2.59E-04     | 3.17E-04 | 3.06E-04  | **1.30E-04** |
-|                                    | avg_psnr | 35.91    | 36.21        | 35.88    | 36.09     | **40.17**    |
-|                                    | avg_ssim | 95.19%   | 95.18%       | 95.00%   | 95.05%    | **97.73%**   |
-| aliberts/paris_street              | avg_mse  | 6.89E-04 | 6.70E-04     | 4.03E-03 | 4.02E-03  | **3.09E-04** |
-|                                    | avg_psnr | 33.48    | 33.68        | 32.05    | 32.15     | **35.40**    |
-|                                    | avg_ssim | 93.76%   | 93.75%       | 89.46%   | 89.46%    | **95.46%**   |
-| aliberts/kitchen                   | avg_mse  | 2.50E-04 | 2.24E-04     | 4.28E-04 | 4.18E-04  | **1.53E-04** |
-|                                    | avg_psnr | 36.73    | 37.33        | 36.56    | 36.75     | **39.12**    |
-|                                    | avg_ssim | 95.47%   | 95.58%       | 95.52%   | 95.53%    | **96.82%**   |
+|                                   |          | vcodec   | pix_fmt      |          |           |              |
+| --------------------------------- | -------- | -------- | ------------ | -------- | --------- | ------------ |
+|                                   |          | libx264  |              | libx265  |           | libsvtav1    |
+| repo_id                           | metric   | yuv420p  | yuv444p      | yuv420p  | yuv444p   | yuv420p      |
+| lerobot/pusht_image               | avg_mse  | 2.90E-04 | **2.03E-04** | 3.13E-04 | 2.29E-04  | 2.19E-04     |
+|                                   | avg_psnr | 35.44    | 37.07        | 35.49    | **37.30** | 37.20        |
+|                                   | avg_ssim | 98.28%   | **98.85%**   | 98.31%   | 98.84%    | 98.72%       |
+| lerobot/aloha_mobile_shrimp_image | avg_mse  | 2.76E-04 | 2.59E-04     | 3.17E-04 | 3.06E-04  | **1.30E-04** |
+|                                   | avg_psnr | 35.91    | 36.21        | 35.88    | 36.09     | **40.17**    |
+|                                   | avg_ssim | 95.19%   | 95.18%       | 95.00%   | 95.05%    | **97.73%**   |
+| lerobot/paris_street              | avg_mse  | 6.89E-04 | 6.70E-04     | 4.03E-03 | 4.02E-03  | **3.09E-04** |
+|                                   | avg_psnr | 33.48    | 33.68        | 32.05    | 32.15     | **35.40**    |
+|                                   | avg_ssim | 93.76%   | 93.75%       | 89.46%   | 89.46%    | **95.46%**   |
+| lerobot/kitchen                   | avg_mse  | 2.50E-04 | 2.24E-04     | 4.28E-04 | 4.18E-04  | **1.53E-04** |
+|                                   | avg_psnr | 36.73    | 37.33        | 36.56    | 36.75     | **39.12**    |
+|                                   | avg_ssim | 95.47%   | 95.58%       | 95.52%   | 95.53%    | **96.82%**   |

docker/Dockerfile.internal

@@ -24,7 +24,7 @@ ARG OS_VERSION=22.04
 FROM nvidia/cuda:${CUDA_VERSION}-base-ubuntu${OS_VERSION}
 
 # Define Python version argument
-ARG PYTHON_VERSION=3.10
+ARG PYTHON_VERSION=3.12
 
 # Configure environment variables
 ENV DEBIAN_FRONTEND=noninteractive \
@@ -85,6 +85,8 @@ RUN if [ "$UNBOUND_DEPS" = "true" ]; then \
 
 RUN uv pip install --no-cache ".[all]"
 
+RUN chmod +x /lerobot/.venv/lib/python${PYTHON_VERSION}/site-packages/triton/backends/nvidia/bin/ptxas
+
 # Copy the rest of the application source code
 # Make sure to have the git-LFS files for testing
 COPY --chown=user_lerobot:user_lerobot . .

docker/Dockerfile.user

@@ -18,8 +18,10 @@
 # docker build -f docker/Dockerfile.user -t lerobot-user .
 # docker run -it --rm lerobot-user
 
+# With USB physical access : docker run -it --device=/dev/ -v /dev/:/dev/ --rm lerobot-user
+
 # Configure the base image
-ARG PYTHON_VERSION=3.10
+ARG PYTHON_VERSION=3.12
 FROM python:${PYTHON_VERSION}-slim
 
 # Configure environment variables

docs/source/_toctree.yml

@@ -19,6 +19,8 @@
     title: Multi GPU training
   - local: peft_training
     title: Training with PEFT (e.g., LoRA)
+  - local: rename_map
+    title: Using Rename Map and Empty Cameras
   title: "Tutorials"
 - sections:
   - local: lerobot-dataset-v3
@@ -29,6 +31,8 @@
     title: Using the Dataset Tools
   - local: dataset_subtask
     title: Using Subtasks in the Dataset
+  - local: streaming_video_encoding
+    title: Streaming Video Encoding
   title: "Datasets"
 - sections:
   - local: act
@@ -45,6 +49,8 @@
     title: NVIDIA GR00T N1.5
   - local: xvla
     title: X-VLA
+  - local: multi_task_dit
+    title: Multitask DiT Policy
   - local: walloss
     title: WALL-OSS
   title: "Policies"
@@ -81,6 +87,8 @@
     title: Processors for Robots and Teleoperators
   - local: env_processor
     title: Environment Processors
+  - local: action_representations
+    title: Action Representations
   title: "Robot Processors"
 - sections:
   - local: so101

docs/source/act.mdx

@@ -88,5 +88,8 @@ lerobot-record \
   --dataset.repo_id=${HF_USER}/eval_act_your_dataset \
   --dataset.num_episodes=10 \
   --dataset.single_task="Your task description" \
+  --dataset.streaming_encoding=true \
+  --dataset.encoder_threads=2 \
+  # --dataset.vcodec=auto \
   --policy.path=${HF_USER}/act_policy
 ```

docs/source/action_representations.mdx

@@ -0,0 +1,223 @@
+# Action Representations
+
+This guide explains the different ways robot actions can be represented in LeRobot, how they relate to each other, and when to use each one.
+
+## Joint Space vs End-Effector Space
+
+Before discussing action representations, it helps to understand the two coordinate spaces actions can live in.
+
+### Joint Space
+
+Joint-space actions directly specify target positions for each motor. For a 6-DOF arm with a gripper, a joint-space action might look like:
+
+```
+action = [shoulder_pan: 45.0, shoulder_lift: -20.0, elbow: -30.0, wrist_pitch: 10.0, wrist_roll: 0.0, wrist_yaw: 5.0, gripper: 0.8]
+```
+
+Joint space is the default in LeRobot. It is simple, requires no kinematics model, and maps directly to motor commands. Most beginner setups (SO-100, Koch) use joint-space actions.
+
+### End-Effector (EE) Space
+
+End-effector-space actions specify the desired position and orientation of the robot's tool tip (gripper) in Cartesian coordinates:
+
+```
+action = [x: 0.25, y: -0.10, z: 0.15, wx: 0.0, wy: 0.0, wz: 0.1, gripper: 0.8]
+```
+
+EE space is more intuitive for tasks like pick-and-place because it directly describes where the gripper should go, but it requires a kinematics model (URDF) to convert between EE poses and joint angles.
+
+### Converting Between Spaces
+
+LeRobot provides processor steps for converting between joint and EE spaces using forward and inverse kinematics. These are built on top of `RobotKinematics`, which loads a URDF model of your robot.
+
+```python
+from lerobot.model.kinematics import RobotKinematics
+from lerobot.robots.so_follower.robot_kinematic_processor import (
+    ForwardKinematicsJointsToEE,
+    InverseKinematicsEEToJoints,
+)
+
+kinematics = RobotKinematics(
+    urdf_path="./SO101/so101_new_calib.urdf",
+    target_frame_name="gripper_frame_link",
+    joint_names=["shoulder", "elbow", "wrist_pitch", "wrist_roll", "wrist_yaw"],
+)
+
+# Joints → EE (for observations: "where is my gripper?")
+fk_step = ForwardKinematicsJointsToEE(kinematics=kinematics, motor_names=[...])
+
+# EE → Joints (for actions: "move my gripper here")
+ik_step = InverseKinematicsEEToJoints(kinematics=kinematics, motor_names=[...])
+```
+
+See [`examples/so100_to_so100_EE/`](https://github.com/huggingface/lerobot/tree/main/examples/so100_to_so100_EE) for a complete working example of recording, replaying, and evaluating with EE-space actions on an SO-100 arm.
+
+## Absolute, Relative, and Delta Actions
+
+Regardless of whether you work in joint space or EE space, the action values can be expressed in three different ways. The terminology follows [UMI (Chi et al., 2024)](https://arxiv.org/abs/2402.10329).
+
+### Absolute Actions (LeRobot default)
+
+Each action specifies the target position directly.
+
+**Example** (joint space, chunk of 4):
+
+```
+current_state = [45.0, -30.0, 10.0]
+
+action_chunk = [
+    [46.0, -29.0, 11.0],   # go to 46, -29, 11
+    [47.5, -27.0, 12.0],   # go to 47.5, -27, 12
+    [49.0, -25.0, 13.5],   # go to 49, -25, 13.5
+    [50.0, -24.0, 15.0],   # go to 50, -24, 15
+]
+```
+
+Each value is a target position in the robot's coordinate frame. Simple and direct, but requires a consistent global coordinate frame. This is the default in LeRobot.
+
+### Relative Actions (used by OpenPI / pi0)
+
+Each action in the chunk is an offset from the **current state at the moment of prediction**. All actions in the chunk share the same reference point:
+
+```
+current_state = [45.0, -30.0, 10.0]
+
+relative_chunk = [
+    [1.0,  1.0, 1.0],   # +1 from current → target 46, -29, 11
+    [2.5,  3.0, 2.0],   # +2.5 from current → target 47.5, -27, 12
+    [4.0,  5.0, 3.5],   # +4 from current → target 49, -25, 13.5
+    [5.0,  6.0, 5.0],   # +5 from current → target 50, -24, 15
+]
+```
+
+The conversion is straightforward: `relative = absolute - current_state`. To recover absolute: `absolute = relative + current_state`.
+
+**Why use relative actions?** The model learns to predict offsets centered around zero, which is easier to normalize and leads to more stable training. Because every chunk references the same current state, there is no error accumulation across chunks.
+
+### Delta Actions (sequential differences)
+
+Each action is an offset from the **previous action** (or from the current state for the first step):
+
+```
+current_state = [45.0, -30.0, 10.0]
+
+delta_chunk = [
+    [1.0,  1.0, 1.0],   # current → 46, -29, 11
+    [1.5,  2.0, 1.0],   # previous action → 47.5, -27, 12
+    [1.5,  2.0, 1.5],   # previous action → 49, -25, 13.5
+    [1.0,  1.0, 1.5],   # previous action → 50, -24, 15
+]
+```
+
+Here each step is relative to the one before it. To recover absolute positions you must sum all previous deltas, which means errors accumulate over time. UMI explicitly argues against this representation for this reason.
+
+### Visual Comparison
+
+The figure below (based on a figure from [UMI, Chi et al., 2024](https://arxiv.org/abs/2402.10329)) illustrates the key difference. With **relative trajectory**, every action in the chunk points back to the same origin (current state), so a new inference step cleanly resets the reference. With **delta**, each action depends on the previous one, so errors accumulate. **Absolute** actions require a consistent global coordinate frame.
+
+<img
+  src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/action_representations_umi.png"
+  alt="Relative Trajectory as Action Representation (UMI, Chi et al., 2024)"
+  width="85%"
+/>
+
+## Using Relative Actions in LeRobot
+
+LeRobot provides `RelativeActionsProcessorStep` to convert between absolute and relative actions inside the processor pipeline. This is how pi0, pi0.5, and pi0_fast support relative actions.
+
+> **Note:** All pi models (pi0, pi0.5, pi0*fast) apply relative conversion \_before* normalization (`relative → normalize`), so the normalizer always sees delta (relative) values. This means **relative action stats are required** for all of them when training with `use_relative_actions=true`. In pi0_fast the `RelativeActionsProcessorStep` only modifies the action — the state observation is unchanged — so `NormalizerProcessorStep` still runs before the state tokenizer and the tokenizer continues to receive normalized state as expected.
+
+### How it works
+
+During **training** (preprocessing), actions are converted from absolute to relative before the model sees them:
+
+```
+raw absolute action → RelativeActionsProcessorStep → normalize → model
+```
+
+During **inference** (postprocessing), model predictions are converted back to absolute before being sent to the robot:
+
+```
+model output → unnormalize → AbsoluteActionsProcessorStep → robot
+```
+
+The `AbsoluteActionsProcessorStep` reads the cached current state from its paired `RelativeActionsProcessorStep`, so the two must be wired together (handled automatically by the policy factory).
+
+### Enabling relative actions for the pi family (pi0, pi0.5, pi0_fast)
+
+**Step 1**: Precompute relative action statistics for your dataset:
+
+```bash
+lerobot-edit-dataset \
+    --repo_id your_dataset \
+    --operation.type recompute_stats \
+    --operation.relative_action true \
+    --operation.chunk_size 50 \
+    --operation.relative_exclude_joints "['gripper']"
+```
+
+**Step 2**: Train with relative actions enabled:
+
+```bash
+lerobot-train \
+    --dataset.repo_id=your_dataset \
+    --policy.type=pi0 \
+    --policy.use_relative_actions=true \
+    --policy.relative_exclude_joints='["gripper"]'
+```
+
+The `relative_exclude_joints` parameter specifies joints that should remain in absolute space. For example, gripper commands are typically binary (open/close) and don't benefit from relative encoding.
+
+### Combining relative actions with RTC
+
+[RTC](https://arxiv.org/abs/2506.07339) runs policy inference at high frequency and sends actions to the robot as they are predicted rather than waiting for a full chunk. Relative actions and RTC are fully compatible: because every chunk in relative mode references the **same** current state (captured at the start of inference), each predicted action in the chunk remains a valid offset even if the robot has already moved. No special handling is needed — `RelativeActionsProcessorStep` caches the state once per inference call and `AbsoluteActionsProcessorStep` applies it to every action in the streamed output.
+
+### Combining relative actions with EE space
+
+Relative actions work in both joint space and EE space. For example, if your dataset stores EE actions, relative encoding converts them to offsets from the current EE pose:
+
+```
+current_ee_state = [x: 0.25, y: -0.10, z: 0.15, gripper: 0.8]
+
+absolute_ee_chunk = [
+    [0.26, -0.09, 0.16, 0.8],
+    [0.28, -0.07, 0.18, 0.8],
+]
+
+relative_ee_chunk = [
+    [0.01,  0.01, 0.01, 0.0],   # offset from current EE pose
+    [0.03,  0.03, 0.03, 0.0],   # offset from current EE pose
+]
+```
+
+## Processing Pipeline Summary
+
+Here is how the different processors compose. Each arrow is a processor step, and they can be chained in a `RobotProcessorPipeline` or `PolicyProcessorPipeline`:
+
+```
+                    ┌─────────────────────────────────────────┐
+   Action Space     │   Joint Space  ←──IK──→  EE Space      │
+                    │   ForwardKinematicsJointsToEE           │
+                    │   InverseKinematicsEEToJoints           │
+                    └─────────────────────────────────────────┘
+
+                    ┌─────────────────────────────────────────┐
+   Representation   │   Absolute  ←────→  Relative            │
+                    │   RelativeActionsProcessorStep (pre)    │
+                    │   AbsoluteActionsProcessorStep (post)   │
+                    └─────────────────────────────────────────┘
+
+                    ┌─────────────────────────────────────────┐
+   Normalization    │   Raw  ←────→  Normalized               │
+                    │   NormalizerProcessorStep (pre)         │
+                    │   UnnormalizerProcessorStep (post)      │
+                    └─────────────────────────────────────────┘
+```
+
+A typical training preprocessor might chain: `raw absolute joint actions → relative → normalize`. A typical inference postprocessor: `unnormalize → absolute → (optionally IK to joints)`.
+
+## References
+
+- [Universal Manipulation Interface (UMI)](https://arxiv.org/abs/2402.10329) - Chi et al., 2024. Defines the relative trajectory action representation and compares it with absolute and delta actions.
+- [Introduction to Processors](./introduction_processors) - How processor pipelines work in LeRobot.
+- [`examples/so100_to_so100_EE/`](https://github.com/huggingface/lerobot/tree/main/examples/so100_to_so100_EE) - Complete example of recording and evaluating with EE-space actions.

docs/source/async.mdx

@@ -48,7 +48,7 @@ python -m lerobot.async_inference.robot_client \
     --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
+    --policy_device=mps \ # POLICY: the device to run the policy on, on the server (cuda, mps, xpu, cpu)
     --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
@@ -310,4 +310,4 @@ Asynchronous inference represents a significant advancement in real-time robotic
 - **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).
+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/huggingface/lerobot/issues).

docs/source/bring_your_own_policies.mdx

@@ -32,7 +32,7 @@ version = "0.1.0"
 dependencies = [
     # your policy-specific dependencies
 ]
-requires-python = ">= 3.11"
+requires-python = ">= 3.12"
 
 [build-system]
 build-backend = # your-build-backend
@@ -41,13 +41,15 @@ requires = # your-build-system
 
 ## Step 2: Define the Policy Configuration
 
-Create a configuration class that inherits from `PreTrainedConfig` and registers your policy type:
+Create a configuration class that inherits from [`PreTrainedConfig`](https://github.com/huggingface/lerobot/blob/main/src/lerobot/configs/policies.py) and registers your policy type:
+Here is a template to get you started, customize the parameters and methods as needed for your policy's architecture and training requirements.
 
 ```python
 # configuration_my_custom_policy.py
 from dataclasses import dataclass, field
 from lerobot.configs.policies import PreTrainedConfig
-from lerobot.configs.types import NormalizationMode
+from lerobot.optim.optimizers import AdamWConfig
+from lerobot.optim.schedulers import CosineDecayWithWarmupSchedulerConfig
 
 @PreTrainedConfig.register_subclass("my_custom_policy")
 @dataclass
@@ -61,62 +63,132 @@ class MyCustomPolicyConfig(PreTrainedConfig):
         hidden_dim: Hidden dimension for the policy network
         # Add your policy-specific parameters here
     """
-    # ...PreTrainedConfig fields...
-    pass
+
+    horizon: int = 50
+    n_action_steps: int = 50
+    hidden_dim: int = 256
+
+    optimizer_lr: float = 1e-4
+    optimizer_weight_decay: float = 1e-4
 
     def __post_init__(self):
         super().__post_init__()
-        # Add any validation logic here
+        if self.n_action_steps > self.horizon:
+            raise ValueError("n_action_steps cannot exceed horizon")
 
     def validate_features(self) -> None:
         """Validate input/output feature compatibility."""
-        # Implement validation logic for your policy's requirements
-        pass
+        if not self.image_features:
+            raise ValueError("MyCustomPolicy requires at least one image feature.")
+        if self.action_feature is None:
+            raise ValueError("MyCustomPolicy requires 'action' in output_features.")
+
+    def get_optimizer_preset(self) -> AdamWConfig:
+        return AdamWConfig(lr=self.optimizer_lr, weight_decay=self.optimizer_weight_decay)
+
+    def get_scheduler_preset(self):
+        return None
+
+    @property
+    def observation_delta_indices(self) -> list[int] | None:
+        """Relative timestep offsets the dataset loader provides per observation.
+
+        Return `None` for single-frame policies. For temporal policies that consume
+        multiple past or future frames, return a list of offsets, e.g. `[-20, -10, 0, 10]` for
+        3 past frames at stride 10 and 1 future frame at stride 10.
+        """
+        return None
+
+    @property
+    def action_delta_indices(self) -> list[int]:
+        """Relative timestep offsets for the action chunk the dataset loader returns.
+        """
+        return list(range(self.horizon))
+
+    @property
+    def reward_delta_indices(self) -> None:
+        return None
 ```
 
 ## Step 3: Implement the Policy Class
 
-Create your policy implementation by inheriting from LeRobot's base `PreTrainedPolicy` class:
+Create your policy implementation by inheriting from [`PreTrainedPolicy`](https://github.com/huggingface/lerobot/blob/main/src/lerobot/policies/pretrained.py):
 
 ```python
 # modeling_my_custom_policy.py
 import torch
 import torch.nn as nn
-from typing import Dict, Any
+from typing import Any
 
 from lerobot.policies.pretrained import PreTrainedPolicy
+from lerobot.utils.constants import ACTION
 from .configuration_my_custom_policy import MyCustomPolicyConfig
 
 class MyCustomPolicy(PreTrainedPolicy):
-    config_class = MyCustomPolicyConfig
+    config_class = MyCustomPolicyConfig  # must match the string in @register_subclass
     name = "my_custom_policy"
 
-    def __init__(self, config: MyCustomPolicyConfig, dataset_stats: Dict[str, Any] = None):
+    def __init__(self, config: MyCustomPolicyConfig, dataset_stats: dict[str, Any] = None):
         super().__init__(config, dataset_stats)
+        config.validate_features()  # not called automatically by the base class
+        self.config = config
+        self.model = ...  # your nn.Module here
+
+    def reset(self):
+        """Reset episode state."""
         ...
+
+    def get_optim_params(self) -> dict:
+        """Return parameters to pass to the optimizer (e.g. with per-group lr/wd)."""
+        return {"params": self.parameters()}
+
+    def predict_action_chunk(self, batch: dict[str, torch.Tensor], **kwargs) -> torch.Tensor:
+        """Return the full action chunk (B, chunk_size, action_dim) for the current observation."""
+        ...
+
+    def select_action(self, batch: dict[str, torch.Tensor], **kwargs) -> torch.Tensor:
+        """Return a single action for the current timestep (called at inference)."""
+        ...
+
+    def forward(self, batch: dict[str, torch.Tensor]) -> dict[str, torch.Tensor]:
+        """Compute the training loss.
+
+        `batch["action_is_pad"]` is a bool mask of shape (B, horizon) that marks
+        timesteps padded because the episode ended before `horizon` steps, you
+        can exclude those from your loss.
+        """
+        actions = batch[ACTION]
+        action_is_pad = batch.get("action_is_pad")
+        ...
+        return {"loss": ...}
 ```
 
 ## Step 4: Add Data Processors
 
-Create processor functions:
+Create processor functions. For a concrete reference, see [processor_act.py](https://github.com/huggingface/lerobot/blob/main/src/lerobot/policies/act/processor_act.py) or [processor_diffusion.py](https://github.com/huggingface/lerobot/blob/main/src/lerobot/policies/diffusion/processor_diffusion.py).
 
 ```python
 # processor_my_custom_policy.py
-from typing import Dict, Any
+from typing import Any
 import torch
 
+from lerobot.processor import PolicyAction, PolicyProcessorPipeline
+
 
 def make_my_custom_policy_pre_post_processors(
     config,
+    dataset_stats: dict[str, dict[str, torch.Tensor]] | None = None,
 ) -> tuple[
     PolicyProcessorPipeline[dict[str, Any], dict[str, Any]],
     PolicyProcessorPipeline[PolicyAction, PolicyAction],
 ]:
-    """Create preprocessing and postprocessing functions for your policy."""
-    pass  # Define your preprocessing and postprocessing logic here
-
+    preprocessor = ...   # build your PolicyProcessorPipeline for inputs
+    postprocessor = ...  # build your PolicyProcessorPipeline for outputs
+    return preprocessor, postprocessor
 ```
 
+**Important - function naming:** LeRobot discovers your processor by name. The function **must** be called `make_{policy_name}_pre_post_processors` (matching the string you passed to `@PreTrainedConfig.register_subclass`).
+
 ## Step 5: Package Initialization
 
 Expose your classes in the package's `__init__.py`:

docs/source/earthrover_mini_plus.mdx

@@ -13,7 +13,7 @@ The EarthRover Mini Plus is a fully open source mobile robot that connects throu
 ### Hardware
 
 - EarthRover Mini robot
-- Computer with Python 3.10 or newer
+- Computer with Python 3.12 or newer
 - Internet connection
 
 ### Setting Up the Frodobots SDK
@@ -170,13 +170,13 @@ Once you can drive the robot well, you can start recording data to train AI mode
 We use Hugging Face to store your data online. First, log in with your token from [Hugging Face settings](https://huggingface.co/settings/tokens):
 
 ```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
+hf auth login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
 ```
 
 Store your Hugging Face username:
 
 ```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
+HF_USER=$(hf auth whoami | awk -F': *' 'NR==1 {print $2}')
 echo $HF_USER
 ```
 
@@ -185,13 +185,16 @@ echo $HF_USER
 Use the standard recording command:
 
 ```bash
-python src/lerobot/scripts/lerobot_record.py \
+lerobot-record \
     --robot.type=earthrover_mini_plus \
     --teleop.type=keyboard_rover \
     --dataset.repo_id=your_username/dataset_name \
     --dataset.num_episodes=2 \
     --dataset.fps=10 \
     --dataset.single_task="Navigate around obstacles" \
+    --dataset.streaming_encoding=true \
+    --dataset.encoder_threads=2 \
+    # --dataset.vcodec=auto \
     --display_data=true
 ```
 
@@ -201,22 +204,26 @@ Replace `your_username/dataset_name` with your Hugging Face username and a name
 
 Your dataset includes:
 
-**Your Actions (2 things)**:
+**Your Actions (2 features)**:
 
-- How much you moved forward/backward
-- How much you turned left/right
+- `linear_velocity`: How much you moved forward/backward
+- `angular_velocity`: How much you turned left/right
 
-**Robot Observations (12 things)**:
+**Robot Observations (24 features)**:
 
 - Front camera video
 - Rear camera video
 - Current speed
 - Battery level
-- Which way the robot is facing
-- GPS location (latitude, longitude, signal strength)
+- Orientation
+- GPS (latitude, longitude, signal strength)
 - Network signal strength
 - Vibration level
-- Lamp status (on/off)
+- Lamp state (on/off)
+- Accelerometer (x, y, z)
+- Gyroscope (x, y, z)
+- Magnetometer (x, y, z)
+- Wheel RPMs (4 wheels)
 
 ### Where Your Data Goes
 

docs/source/envhub.mdx

@@ -155,10 +155,10 @@ Upload your repository to Hugging Face:
 pip install huggingface_hub
 
 # Login to Hugging Face
-huggingface-cli login
+hf auth login
 
 # Create a new repository
-huggingface-cli repo create my-custom-env --type space --org my-org
+hf repo create my-org/my-custom-env
 
 # Initialize git and push
 git init

docs/source/groot.mdx

@@ -120,9 +120,12 @@ lerobot-record \
   --display_data=true \
   --dataset.repo_id=<user>/eval_groot-bimanual  \
   --dataset.num_episodes=10 \
-  --dataset.single_task="Grab and handover the red cube to the other arm"
-  --policy.path=<user>/groot-bimanual # your trained model
-  --dataset.episode_time_s=30
+  --dataset.single_task="Grab and handover the red cube to the other arm" \
+  --dataset.streaming_encoding=true \
+  --dataset.encoder_threads=2 \
+  # --dataset.vcodec=auto \
+  --policy.path=<user>/groot-bimanual \ # your trained model
+  --dataset.episode_time_s=30 \
   --dataset.reset_time_s=10
 ```
 

docs/source/hope_jr.mdx

@@ -224,12 +224,15 @@ lerobot-record \
     --teleop.port=/dev/tty.usbmodem1201 \
     --teleop.id=right \
     --teleop.side=right \
-    --dataset.repo_id=nepyope/hand_record_test_with_video_data \
+    --dataset.repo_id=<USER>/hand_record_test_with_video_data \
     --dataset.single_task="Hand recording test with video data" \
     --dataset.num_episodes=1 \
     --dataset.episode_time_s=5 \
     --dataset.push_to_hub=true \
     --dataset.private=true \
+    --dataset.streaming_encoding=true \
+    --dataset.encoder_threads=2 \
+    # --dataset.vcodec=auto \
     --display_data=true
 ```
 
@@ -241,7 +244,7 @@ lerobot-replay \
     --robot.port=/dev/tty.usbmodem58760432281 \
     --robot.id=right \
     --robot.side=right \
-    --dataset.repo_id=nepyope/hand_record_test_with_camera \
+    --dataset.repo_id=<USER>/hand_record_test_with_camera \
     --dataset.episode=0
 ```
 
@@ -249,13 +252,13 @@ lerobot-replay \
 
 ```bash
 lerobot-train \
-  --dataset.repo_id=nepyope/hand_record_test_with_video_data \
+  --dataset.repo_id=<USER>/hand_record_test_with_video_data \
   --policy.type=act \
   --output_dir=outputs/train/hopejr_hand \
   --job_name=hopejr \
   --policy.device=mps \
   --wandb.enable=true \
-  --policy.repo_id=nepyope/hand_test_policy
+  --policy.repo_id=<USER>/hand_test_policy
 ```
 
 ### Evaluate
@@ -270,8 +273,11 @@ lerobot-record \
   --robot.side=right \
   --robot.cameras='{"main": {"type": "opencv", "index_or_path": 0, "width": 640, "height": 480, "fps": 30}}' \
   --display_data=false \
-  --dataset.repo_id=nepyope/eval_hopejr \
+  --dataset.repo_id=<USER>/eval_hopejr \
   --dataset.single_task="Evaluate hopejr hand policy" \
   --dataset.num_episodes=10 \
+  --dataset.streaming_encoding=true \
+  --dataset.encoder_threads=2 \
+  # --dataset.vcodec=auto \
   --policy.path=outputs/train/hopejr_hand/checkpoints/last/pretrained_model
 ```

docs/source/il_robots.mdx

@@ -159,13 +159,13 @@ We use the Hugging Face hub features for uploading your dataset. If you haven't
 Add your token to the CLI by running this command:
 
 ```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
+hf auth login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
 ```
 
 Then store your Hugging Face repository name in a variable:
 
 ```bash
-HF_USER=$(hf auth whoami | head -n 1)
+HF_USER=$(NO_COLOR=1 hf auth whoami | awk -F': *' 'NR==1 {print $2}')
 echo $HF_USER
 ```
 
@@ -185,7 +185,10 @@ lerobot-record \
     --display_data=true \
     --dataset.repo_id=${HF_USER}/record-test \
     --dataset.num_episodes=5 \
-    --dataset.single_task="Grab the black cube"
+    --dataset.single_task="Grab the black cube" \
+    --dataset.streaming_encoding=true \
+    # --dataset.vcodec=auto \
+    --dataset.encoder_threads=2
 ```
 </hfoption>
 <hfoption id="API example">
@@ -324,7 +327,7 @@ You can look for other LeRobot datasets on the hub by searching for `LeRobot` [t
 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
+hf upload ${HF_USER}/record-test ~/.cache/huggingface/lerobot/{repo-id} --repo-type dataset
 ```
 
 #### Record function
@@ -421,7 +424,7 @@ robot = SO100Follower(robot_config)
 robot.connect()
 
 dataset = LeRobotDataset("<hf_username>/<dataset_repo_id>", episodes=[episode_idx])
-actions = dataset.hf_dataset.select_columns("action")
+actions = dataset.select_columns("action")
 
 log_say(f"Replaying episode {episode_idx}")
 for idx in range(dataset.num_frames):
@@ -488,7 +491,7 @@ If your local computer doesn't have a powerful GPU you could utilize Google Cola
 Once training is done, upload the latest checkpoint with:
 
 ```bash
-huggingface-cli upload ${HF_USER}/act_so101_test \
+hf upload ${HF_USER}/act_so101_test \
   outputs/train/act_so101_test/checkpoints/last/pretrained_model
 ```
 
@@ -496,7 +499,7 @@ You can also upload intermediate checkpoints with:
 
 ```bash
 CKPT=010000
-huggingface-cli upload ${HF_USER}/act_so101_test${CKPT} \
+hf upload ${HF_USER}/act_so101_test${CKPT} \
   outputs/train/act_so101_test/checkpoints/${CKPT}/pretrained_model
 ```
 
@@ -515,6 +518,9 @@ lerobot-record  \
   --display_data=false \
   --dataset.repo_id=${HF_USER}/eval_so100 \
   --dataset.single_task="Put lego brick into the transparent box" \
+  --dataset.streaming_encoding=true \
+  --dataset.encoder_threads=2 \
+  # --dataset.vcodec=auto \
   # <- Teleop optional if you want to teleoperate in between episodes \
   # --teleop.type=so100_leader \
   # --teleop.port=/dev/ttyACM0 \

docs/source/installation.mdx

@@ -1,8 +1,8 @@
 # Installation
 
-This guide uses conda (via miniforge) to manage environments. If you prefer another environment manager (e.g. `uv`, `venv`), ensure you have Python >=3.10 and ffmpeg installed with the `libsvtav1` encoder, then skip ahead to [Install LeRobot](#step-3-install-lerobot-).
+This guide uses `conda` (via miniforge) to manage environments (recommended). If you prefer another environment manager (e.g. `uv`, `venv`), ensure you have Python >=3.12 and `ffmpeg` installed with the `libsvtav1` encoder, then skip ahead to [Environment Setup](#step-2-environment-setup).
 
-## Step 1: Install [`miniforge`](https://conda-forge.org/download/)
+## Step 1 (`conda` only): Install [`miniforge`](https://conda-forge.org/download/)
 
 ```bash
 wget "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
@@ -11,22 +11,47 @@ bash Miniforge3-$(uname)-$(uname -m).sh
 
 ## Step 2: Environment Setup
 
-Create a virtual environment with Python 3.10, using conda:
+Create a virtual environment with Python 3.12:
 
+<!-- prettier-ignore-start -->
+<hfoptions id="create_venv">
+<hfoption id="conda">
 ```bash
-conda create -y -n lerobot python=3.10
+conda create -y -n lerobot python=3.12
 ```
-
-Then activate your conda environment, you have to do this each time you open a shell to use lerobot:
-
+</hfoption>
+<hfoption id="uv">
 ```bash
+uv python install 3.12
+uv venv --python 3.12
+```
+</hfoption>
+</hfoptions>
+<!-- prettier-ignore-end -->
+
+Then activate your virtual environment, you have to do this each time you open a shell to use lerobot:
+
+<!-- prettier-ignore-start -->
+<hfoptions id="activate_venv">
+<hfoption id="conda">```bash
 conda activate lerobot
+```</hfoption>
+<hfoption id="uv">
+```bash
+# Linux/macOSsource
+source .venv/bin/activate
+# Windows PowerShell
+source .venv\Scripts\Activate.ps1
 ```
+</hfoption>
+</hfoptions>
+<!-- prettier-ignore-end -->
 
 When using `conda`, install `ffmpeg` in your environment:
 
 ```bash
 conda install ffmpeg -c conda-forge
+ffmpeg -version  # ffmpeg 8.X is not yet supported !
 ```
 
 > [!TIP]
@@ -40,6 +65,16 @@ conda install ffmpeg -c conda-forge
 >
 > - _[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`.
 
+> [!NOTE]
+> When installing LeRobot inside WSL (Windows Subsystem for Linux), make sure to install `evdev` with the following command:
+>
+> ```bash
+> conda install evdev -c conda-forge
+> ```
+
+> [!IMPORTANT]
+> If you are using `uv` you will have to install `ffmpeg` system-wide (outside of the virtual environment). You rely on `uv` and `torchcodec` ability to dynamically link to the system `ffmpeg`.
+
 ## Step 3: Install LeRobot 🤗
 
 ### From Source
@@ -53,23 +88,45 @@ cd lerobot
 
 Then, install the library in editable mode. This is useful if you plan to contribute to the code.
 
+<!-- prettier-ignore-start -->
+<hfoptions id="install_lerobot_src">
+<hfoption id="conda">
 ```bash
 pip install -e .
 ```
+</hfoption>
+<hfoption id="uv">
+```bash
+uv pip install -e .
+```
+</hfoption>
+</hfoptions>
+<!-- prettier-ignore-end -->
 
 ### Installation from PyPI
 
 **Core Library:**
 Install the base package with:
 
+<!-- prettier-ignore-start -->
+<hfoptions id="install_lerobot_pypi">
+<hfoption id="conda">
 ```bash
 pip install lerobot
 ```
+</hfoption>
+<hfoption id="uv">
+```bash
+uv pip install lerobot
+```
+</hfoption>
+</hfoptions>
+<!-- prettier-ignore-end -->
 
 _This installs only the default dependencies._
 
 **Extra Features:**
-To install additional functionality, use one of the following:
+To install additional functionality, use one of the following (If you are using `uv`, replace `pip install` with `uv pip install` in the commands below.):
 
 ```bash
 pip install 'lerobot[all]'          # All available features
@@ -83,13 +140,10 @@ _Replace `[...]` with your desired features._
 For a full list of optional dependencies, see:
 https://pypi.org/project/lerobot/
 
-> [!NOTE]
-> For lerobot 0.4.0, if you want to install pi, you will have to do: `pip install "lerobot[pi]@git+https://github.com/huggingface/lerobot.git"`
-
 ### 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:
+To install these for Linux run:
 
 ```bash
 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
@@ -99,7 +153,7 @@ For other systems, see: [Compiling PyAV](https://pyav.org/docs/develop/overview/
 
 ## Optional dependencies
 
-LeRobot provides optional extras for specific functionalities. Multiple extras can be combined (e.g., `.[aloha,feetech]`). For all available extras, refer to `pyproject.toml`.
+LeRobot provides optional extras for specific functionalities. Multiple extras can be combined (e.g., `.[aloha,feetech]`). For all available extras, refer to `pyproject.toml`. If you are using `uv`, replace `pip install` with `uv pip install` in the commands below.
 
 ### Simulations
 

docs/source/lekiwi.mdx

@@ -279,13 +279,13 @@ We use the Hugging Face hub features for uploading your dataset. If you haven't
 Add your token to the CLI by running this command:
 
 ```bash
-huggingface-cli login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
+hf auth login --token ${HUGGINGFACE_TOKEN} --add-to-git-credential
 ```
 
 Then store your Hugging Face repository name in a variable:
 
 ```bash
-HF_USER=$(huggingface-cli whoami | head -n 1)
+HF_USER=$(hf auth whoami | awk -F': *' 'NR==1 {print $2}')
 echo $HF_USER
 ```
 

docs/source/lerobot-dataset-v3.mdx

@@ -41,7 +41,10 @@ lerobot-record \
   --display_data=true \
   --dataset.repo_id=${HF_USER}/record-test \
   --dataset.num_episodes=5 \
-  --dataset.single_task="Grab the black cube"
+  --dataset.single_task="Grab the black cube" \
+  --dataset.streaming_encoding=true \
+  # --dataset.vcodec=auto \
+  --dataset.encoder_threads=2
 ```
 
 See the [recording guide](./il_robots#record-a-dataset) for more details.

docs/source/multi_task_dit.mdx

@@ -0,0 +1,340 @@
+# Multitask DiT Policy
+
+Multitask Diffusion Transformer (DiT) Policy is an evolution of the original Diffusion Policy architecture, which leverages a large DiT with text and vision conditioning for multitask robot learning. This implementation supports both diffusion and flow matching objectives for action generation, enabling robots to perform diverse manipulation tasks conditioned on language instructions.
+
+## Model Overview
+
+The model uses:
+
+- **CLIP Vision Encoder**: Processes RGB images from multiple camera views
+- **CLIP Text Encoder**: Encodes language task instructions (frozen weights with learnable projection)
+- **Diffusion Transformer**: Predicts action sequences conditioned on observations and language
+- **Two Objectives**: Supports both diffusion (DDPM/DDIM) and flow matching for action generation
+
+This model is exciting because you can achieve extremely high dexterity, competitive with multi-billion parameter
+VLAs, with only ~450M parameters and significantly less training.
+
+## Installation Requirements
+
+Multitask DiT Policy has additional dependencies. Install it with:
+
+```bash
+pip install lerobot[multi_task_dit]
+```
+
+This will install all necessary dependencies including the HuggingFace Transformers library for CLIP models.
+
+## Usage
+
+To use Multitask DiT in your LeRobot configuration, specify the policy type as:
+
+```python
+policy.type=multi_task_dit
+```
+
+## Training
+
+### Basic Training Command
+
+Here's a complete training command for training Multitask DiT on your dataset:
+
+```bash
+lerobot-train \
+  --dataset.repo_id=YOUR_DATASET \
+  --output_dir=./outputs/multitask_dit_training \
+  --batch_size=32 \
+  --steps=5000 \
+  --save_freq=500 \
+  --log_freq=100 \
+  --policy.type=multi_task_dit \
+  --policy.device=cuda \
+  --policy.repo_id="HF_USER/multitask-dit-your-robot" \
+  --wandb.enable=true
+```
+
+### Recommended Hyperparameters and Dataset Details (30Hz Control Frequency)
+
+For reliable performance, start with these suggested default hyperparameters:
+
+```bash
+lerobot-train \
+  --dataset.repo_id=YOUR_DATASET \
+  --output_dir=./outputs/mutitask_dit_training \
+  --batch_size=320 \
+  --steps=30000 \
+  --policy.type=multi_task_dit \
+  --policy.device=cuda \
+  --policy.horizon=32 \
+  --policy.n_action_steps=24 \
+  --policy.objective=diffusion \
+  --policy.noise_scheduler_type=DDPM \
+  --policy.num_train_timesteps=100 \
+  --policy.repo_id="HF_USER/multitask-dit-your-robot" \
+  --wandb.enable=true
+```
+
+**Key Parameters:**
+
+- **Batch Size**: 192-320 - If you have access to a GPU that can support this, you will get the best training dynamics
+- **Horizon**: 32 - number of action steps to predict, ~1.0 sec at 30Hz
+- **n_action_steps**: 24 - ~0.8 seconds at 30Hz
+- **Objective**: `diffusion` - start with diffusion and experiment with flow matching if generation quality is poor
+- **Training Steps**: >30k steps recommended for a single task
+
+### Training Configuration Parameters
+
+#### Objective Selection
+
+Choose between diffusion and flow matching:
+
+```bash
+# Diffusion objective (default)
+--policy.objective=diffusion \
+--policy.noise_scheduler_type=DDPM \  # or "DDIM"
+--policy.num_train_timesteps=100 \
+--policy.num_inference_steps=10 \  # For faster inference
+--policy.beta_schedule=squaredcos_cap_v2 \  # Noise schedule type
+--policy.prediction_type=epsilon \  # "epsilon" (predict noise) or "sample" (predict clean)
+--policy.clip_sample=true \  # Clip samples during denoising
+--policy.clip_sample_range=1.0  # Clipping range [-x, x]
+
+# Flow matching objective
+--policy.objective=flow_matching \
+--policy.timestep_sampling_strategy=beta \  # or "uniform" | the beta sampling strategy performance appears much better in practice
+--policy.num_integration_steps=100 \
+--policy.integration_method=euler \  # or "rk4"
+--policy.sigma_min=0.0  # Minimum noise in flow interpolation path
+```
+
+#### Transformer Architecture
+
+Adjust model capacity based on dataset size:
+
+```bash
+# Small datasets (< 100 examples)
+--policy.num_layers=4 \
+--policy.hidden_dim=512 \
+--policy.num_heads=8  # should ideally be hidden_dim // 64
+
+# Medium datasets (100-5k examples) - default
+--policy.num_layers=6 \
+--policy.hidden_dim=512 \
+--policy.num_heads=8  # should ideally be hidden_dim // 64
+
+# Large datasets (> 5k examples)
+--policy.num_layers=8 \
+--policy.hidden_dim=512 \
+--policy.num_heads=8   # should ideally be hidden_dim // 64
+```
+
+**Positional Encoding Options:**
+
+The model supports two positional encoding methods for action sequences:
+
+```bash
+# Rotary Position Embedding (RoPE) - default, recommended
+--policy.use_rope=true \
+--policy.rope_base=10000.0  # Base frequency for RoPE
+
+# Absolute positional encoding
+--policy.use_positional_encoding=true  # Disables RoPE when true
+```
+
+**Other Transformer Parameters:**
+
+```bash
+--policy.dropout=0.1  # Dropout rate for DiT blocks (0.0-1.0)
+--policy.timestep_embed_dim=256  # Timestep embedding dimension
+```
+
+#### Vision Encoder Configuration
+
+```bash
+# Use different CLIP model for more expressivity at the cost of inference time
+# experiment with larger or smaller models depending on the complexity of your tasks and size of dataset
+--policy.vision_encoder_name=openai/clip-vit-large-patch14
+
+# Use separate vision encoder per camera
+# This may be useful when cameras have significantly different characteristics, but
+# be wary of increased VRAM footprint.
+--policy.use_separate_rgb_encoder_per_camera=true
+
+# Image preprocessing
+--policy.image_resize_shape=[XXX,YYY] \ # you may need to resize your images for inference speed ups
+--policy.image_crop_shape=[224,224] \
+--policy.image_crop_is_random=true  # Random during training, center at inference
+```
+
+#### Text Encoder Configuration
+
+```bash
+# Use different CLIP text encoder model
+# same as vision: experiment with larger or smaller models depending on the
+# complexity of your tasks and size of dataset
+--policy.text_encoder_name=openai/clip-vit-large-patch14
+```
+
+#### Learning Rate Configuration
+
+The vision encoder uses a separate learning rate multiplier, where 1/10th is suggested to be the ideal staritng point:
+
+```bash
+--policy.optimizer_lr=2e-5 \
+--policy.vision_encoder_lr_multiplier=0.1  # Vision encoder LR = 0.1 * optimizer_lr
+```
+
+### Training Tuning Guidelines
+
+#### 1. Flow Matching with Beta Sampling
+
+The original diffusion implementation here is based on the work described in [TRI's LBM paper](https://arxiv.org/abs/2507.05331)
+
+Additionally, we have implemented a flow-matching objective, which is described at a high-level in [Boston Dynamics blog post](https://bostondynamics.com/blog/large-behavior-models-atlas-find-new-footing/).
+
+Consider testing the flow-matching objective and evaluating performance differences for your task:
+
+```bash
+--policy.objective=flow_matching \
+--policy.timestep_sampling_strategy=beta \
+--policy.timestep_sampling_alpha=1.5 \
+--policy.timestep_sampling_beta=1.0 \
+--policy.timestep_sampling_s=0.999
+```
+
+This hasn't been shown to be a silver bullet across every user case, but it occasionally results in smoother and more consistent actions.
+
+#### 2. Number of Transformer Layers
+
+Match model capacity to your dataset size:
+
+- **Small datasets** (< 100 examples): Reduce to 4 layers
+- **Large datasets** (> 5k examples): Increase to 8 layers
+
+#### 3. `horizon` Tuning
+
+The model can be sensitive to the horizon you choose. Start with around a 1 second horizon based on your control frequency:
+
+- **30 Hz frequency**: `horizon=30`
+- **10 Hz frequency**: `horizon=10`
+
+Then experiment with increasing from there. The horizon determines how far into the future the model predicts actions.
+
+#### 4. `n_action_steps` Sensitivity
+
+The model can also be very sensitive to `n_action_steps`. Start with it being around 0.8 seconds based on your control frequency and tune from there:
+
+- **Lower values**: More reactive but potentially less stable for long-horizon tasks
+- **Higher values**: Better for long-horizon execution but open-loop failures are limited in their recovery
+
+### Inference Tuning
+
+For faster inference, use DDIM with fewer sampling steps:
+
+```bash
+--policy.noise_scheduler_type=DDIM \
+--policy.num_inference_steps=10
+```
+
+### Resuming Training
+
+To resume training from a checkpoint:
+
+```bash
+lerobot-train \
+  --config_path=./outputs/mutitask_dit_training/checkpoints/last/pretrained_model/train_config.json \
+  --resume=true
+```
+
+The checkpoint directory should contain `model.safetensors` and `config.json` files (saved automatically during training). When resuming, the configuration is loaded from the checkpoint, so you don't need to specify other parameters.
+
+## Common Failure Modes and Debugging
+
+Training these models can be finicky. Here are common failure modes and debugging approaches:
+
+### Idling / No Motion
+
+The model may "collapse" during inference, resulting in static or no motion. This can occur when:
+
+1. **Insufficient training data**: If you only have 20-50 examples, try to roughly double your dataset size. Once you have above 300 examples, if you're still seeing this, the task may be too complex.
+
+2. **Multiple similar tasks**: When your dataset contains multiple similar tasks (e.g., picking up 2 different objects), the model may rely too heavily on language conditioning which might not be rich enough.
+
+**Debugging tips:**
+
+- Increase dataset size (double until you get to over 300 examples)
+- Train for longer, up to 100k steps, even when the loss flatlines
+- Check if the model is receiving proper language instructions or increase diversity of instruction
+
+### Executing the Wrong Task
+
+Sometimes the robot will completely ignore your instruction and perform some other task. This generally only happens if you have trained on multiple tasks.
+
+**Potential causes:**
+
+- Language instruction ambiguity
+- Insufficient task-specific training data
+- Model confusion between similar tasks in the multitask dataset
+
+**Debugging tips:**
+
+- Verify language instruction specificity, especially if descriptions are similar between multiple tasks
+- Check task distribution in your training dataset and add weighting to the failing/ignored task
+- Consider task-specific fine-tuning
+
+### Training Instability
+
+If training loss is unstable or diverging:
+
+- Try adjusting learning rate between `1e-5` and `3e-4`
+- Increase batch size if possible
+- Check that your dataset normalization is correct
+- Verify image preprocessing is working correctly
+
+## Performance Considerations
+
+### GPU Requirements
+
+- **Inference**: At least an RTX 5070 Ti (or equivalent GPU) is recommended for reasonable speed performance
+- **Training**: A GPU with enough VRAM to load batch sizes of >64 is ideal, which will vary depending on the number of image observations, etc
+
+### Batch Size Recommendations
+
+- **Minimum**: 64 (less than this may result in unstable training)
+- **Recommended**: 256-320 (best performance, requires larger GPU)
+
+## Example: Training on Custom Dataset
+
+Here's a complete example training on a custom dataset:
+
+```bash
+lerobot-train \
+  --dataset.repo_id=YOUR_DATASET \
+  --output_dir=./outputs/mutitask_dit_training \
+  --batch_size=320 \
+  --steps=30000 \
+  --save_freq=1000 \
+  --log_freq=100 \
+  --eval_freq=1000 \
+  --policy.type=multi_task_dit \
+  --policy.device=cuda \
+  --policy.horizon=32 \
+  --policy.n_action_steps=24 \
+  --policy.objective=diffusion \
+  --policy.noise_scheduler_type=DDPM \
+  --policy.num_layers=6 \
+  --policy.hidden_dim=512 \
+  --policy.vision_encoder_name=openai/clip-vit-base-patch16 \
+  --policy.image_resize_shape=[320,240] \
+  --policy.image_crop_shape=[224,224] \
+  --policy.repo_id="HF_USER/multitask-dit-your-robot" \
+  --wandb.enable=true \
+  --wandb.project=multitask_dit
+```
+
+## References
+
+For more details on the technical implementation and architecture, see:
+
+- [A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulation](https://arxiv.org/abs/2507.05331)
+- [Large Behavior Models and Atlas Find New Footing](https://bostondynamics.com/blog/large-behavior-models-atlas-find-new-footing/)
+- [Dissecting and Open-Sourcing Multitask Diffusion Transformer Policy](https://brysonkjones.substack.com/p/dissecting-and-open-sourcing-multitask-diffusion-transformer-policy)

docs/source/phone_teleop.mdx

@@ -66,12 +66,13 @@ Run on of the examples scripts to teleoperate, record a dataset, replay a datase
 
 All scripts assume you configured your robot (e.g., SO-100 follower) and set the correct serial port.
 
-Additionally you need to **copy the urdf of the robot to the examples folder**. For the examples in this tutorial (Using SO100/SO101) it is highly recommended to use the urdf in the [SO-ARM100 repo](https://github.com/TheRobotStudio/SO-ARM100/blob/main/Simulation/SO101/so101_new_calib.urdf)
+Additionally you need to **copy the URDF of the robot into the examples folder**. For the examples in this tutorial (using SO100/SO101), copy the `SO101` folder from the [SO-ARM100 repo](https://github.com/TheRobotStudio/SO-ARM100/blob/main/Simulation/SO101) into the `examples/phone_to_so100/` directory, so that the URDF file path becomes `examples/phone_to_so100/SO101/so101_new_calib.urdf`.
 
 - Run this example to teleoperate:
 
   ```bash
-  python examples/phone_to_so100/teleoperate.py
+  cd examples/phone_to_so100
+  python teleoperate.py
   ```
 
 After running the example:
@@ -84,19 +85,22 @@ Additionally you can customize mapping or safety limits by editing the processor
 - Run this example to record a dataset, which saves absolute end effector observations and actions:
 
   ```bash
-  python examples/phone_to_so100/record.py
+  cd examples/phone_to_so100
+  python record.py
   ```
 
 - Run this example to replay recorded episodes:
 
   ```bash
-  python examples/phone_to_so100/replay.py
+  cd examples/phone_to_so100
+  python replay.py
   ```
 
 - Run this example to evaluate a pretrained policy:
 
   ```bash
-  python examples/phone_to_so100/evaluate.py
+  cd examples/phone_to_so100
+  python evaluate.py
   ```
 
 ### Important pipeline steps and options

docs/source/pi0.mdx

@@ -34,11 +34,6 @@ As described by Physical Intelligence, while AI has achieved remarkable success
    pip install -e ".[pi]"
    ```
 
-   > [!NOTE]
-   > For lerobot 0.4.0, if you want to install pi tag, you will have to do: `pip install "lerobot[pi]@git+https://github.com/huggingface/lerobot.git"`.
-   >
-   > This will be solved in the next patch release
-
 ## Training Data and Capabilities
 
 π₀ is trained on the largest robot interaction dataset to date, combining three key data sources:
@@ -60,7 +55,7 @@ policy.type=pi0
 For training π₀, you can use the standard LeRobot training script with the appropriate configuration:
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py \
+lerobot-train \
     --dataset.repo_id=your_dataset \
     --policy.type=pi0 \
     --output_dir=./outputs/pi0_training \
@@ -96,6 +91,46 @@ python src/lerobot/scripts/lerobot_train.py \
 
 **💡 Tip**: Setting `train_expert_only=true` freezes the VLM and trains only the action expert and projections, allowing finetuning with reduced memory usage.
 
+## Relative Actions
+
+By default, π₀ predicts absolute actions. You can enable **relative actions** so the model predicts offsets relative to the current robot state. This can improve training stability for certain setups.
+
+To use relative actions, first recompute your dataset stats in relative space via the CLI:
+
+```bash
+lerobot-edit-dataset \
+    --repo_id your_dataset \
+    --operation.type recompute_stats \
+    --operation.relative_action true \
+    --operation.chunk_size 50 \
+    --operation.relative_exclude_joints "['gripper']" \
+    --push_to_hub true
+```
+
+Or equivalently in Python:
+
+```python
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.dataset_tools import recompute_stats
+
+dataset = LeRobotDataset("your_dataset")
+recompute_stats(dataset, relative_action=True, chunk_size=50, relative_exclude_joints=["gripper"])
+dataset.push_to_hub()
+```
+
+The `chunk_size` should match your policy's `chunk_size` (default 50 for π₀). `relative_exclude_joints` lists joint names that should remain in absolute space (e.g. gripper commands). Use `--push_to_hub true` to upload the updated stats to the Hub.
+
+Then train with relative actions enabled:
+
+```bash
+lerobot-train \
+    --dataset.repo_id=your_dataset \
+    --policy.type=pi0 \
+    --policy.use_relative_actions=true \
+    --policy.relative_exclude_joints='["gripper"]' \
+    ...
+```
+
 ## License
 
 This model follows the **Apache 2.0 License**, consistent with the original [OpenPI repository](https://github.com/Physical-Intelligence/openpi).

docs/source/pi05.mdx

@@ -36,11 +36,6 @@ This diverse training mixture creates a "curriculum" that enables generalization
    pip install -e ".[pi]"
    ```
 
-   > [!NOTE]
-   > For lerobot 0.4.0, if you want to install pi tag, you will have to do: `pip install "lerobot[pi]@git+https://github.com/huggingface/lerobot.git"`.
-   >
-   > This will be solved in the next patch release
-
 ## Usage
 
 To use π₀.₅ in your LeRobot configuration, specify the policy type as:
@@ -56,7 +51,7 @@ policy.type=pi05
 Here's a complete training command for finetuning the base π₀.₅ model on your own dataset:
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py\
+lerobot-train \
     --dataset.repo_id=your_dataset \
     --policy.type=pi05 \
     --output_dir=./outputs/pi05_training \
@@ -102,6 +97,46 @@ python src/lerobot/datasets/v30/augment_dataset_quantile_stats.py \
 
 Or train pi05 with this normalization mapping: `--policy.normalization_mapping='{"ACTION": "MEAN_STD", "STATE": "MEAN_STD", "VISUAL": "IDENTITY"}'`
 
+## Relative Actions
+
+By default, π₀.₅ predicts absolute actions. You can enable **relative actions** so the model predicts offsets relative to the current robot state. This can improve training stability for certain setups.
+
+To use relative actions, first recompute your dataset stats in relative space via the CLI:
+
+```bash
+lerobot-edit-dataset \
+    --repo_id your_dataset \
+    --operation.type recompute_stats \
+    --operation.relative_action true \
+    --operation.chunk_size 50 \
+    --operation.relative_exclude_joints "['gripper']" \
+    --push_to_hub true
+```
+
+Or equivalently in Python:
+
+```python
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.dataset_tools import recompute_stats
+
+dataset = LeRobotDataset("your_dataset")
+recompute_stats(dataset, relative_action=True, chunk_size=50, relative_exclude_joints=["gripper"])
+dataset.push_to_hub()
+```
+
+The `chunk_size` should match your policy's `chunk_size` (default 50 for π₀.₅). `relative_exclude_joints` lists joint names that should remain in absolute space (e.g. gripper commands). Use `--push_to_hub true` to upload the updated stats to the Hub.
+
+Then train with relative actions enabled:
+
+```bash
+lerobot-train \
+    --dataset.repo_id=your_dataset \
+    --policy.type=pi05 \
+    --policy.use_relative_actions=true \
+    --policy.relative_exclude_joints='["gripper"]' \
+    ...
+```
+
 ## Performance Results
 
 ### Libero Benchmark Results

docs/source/pi0fast.mdx

@@ -43,16 +43,11 @@ This approach can transform **any existing VLM** into a VLA by training it to pr
    pip install -e ".[pi]"
    ```
 
-   > [!NOTE]
-   > For lerobot 0.4.0, if you want to install the pi tag, you will have to do: `pip install "lerobot[pi]@git+https://github.com/huggingface/lerobot.git"`.
-   >
-   > This will be solved in the next patch release
-
 ## Training a Custom FAST Tokenizer
 
 You have two options for the FAST tokenizer:
 
-1. **Use the pre-trained tokenizer**: The `physical-intelligence/fast` tokenizer was trained on 1M+ real robot action sequences and works as a general-purpose tokenizer.
+1. **Use the pre-trained tokenizer**: The `lerobot/fast-action-tokenizer` tokenizer was trained on 1M+ real robot action sequences and works as a general-purpose tokenizer.
 
 2. **Train your own tokenizer**: For maximum performance on your specific dataset, you can finetune the tokenizer on your own data.
 
@@ -114,15 +109,15 @@ lerobot-train \
 
 ### Key Training Parameters
 
-| Parameter                              | Description                                        | Default                      |
-| -------------------------------------- | -------------------------------------------------- | ---------------------------- |
-| `--policy.gradient_checkpointing=true` | Reduces memory usage significantly during training | `false`                      |
-| `--policy.dtype=bfloat16`              | Use mixed precision training for efficiency        | `float32`                    |
-| `--policy.chunk_size`                  | Number of action steps to predict (action horizon) | `50`                         |
-| `--policy.n_action_steps`              | Number of action steps to execute                  | `50`                         |
-| `--policy.max_action_tokens`           | Maximum number of FAST tokens per action chunk     | `256`                        |
-| `--policy.action_tokenizer_name`       | FAST tokenizer to use                              | `physical-intelligence/fast` |
-| `--policy.compile_model=true`          | Enable torch.compile for faster training           | `false`                      |
+| Parameter                              | Description                                        | Default                         |
+| -------------------------------------- | -------------------------------------------------- | ------------------------------- |
+| `--policy.gradient_checkpointing=true` | Reduces memory usage significantly during training | `false`                         |
+| `--policy.dtype=bfloat16`              | Use mixed precision training for efficiency        | `float32`                       |
+| `--policy.chunk_size`                  | Number of action steps to predict (action horizon) | `50`                            |
+| `--policy.n_action_steps`              | Number of action steps to execute                  | `50`                            |
+| `--policy.max_action_tokens`           | Maximum number of FAST tokens per action chunk     | `256`                           |
+| `--policy.action_tokenizer_name`       | FAST tokenizer to use                              | `lerobot/fast-action-tokenizer` |
+| `--policy.compile_model=true`          | Enable torch.compile for faster training           | `false`                         |
 
 ## Inference
 

docs/source/policy_multi_task_dit_README.md

@@ -0,0 +1,37 @@
+# Multitask DiT Policy
+
+## Citation
+
+If you use this work, please cite the following works:
+
+```bibtex
+@misc{jones2025multitaskditpolicy,
+  author = {Bryson Jones},
+  title = {Dissecting and Open-Sourcing Multitask Diffusion Transformer Policy},
+  year = {2025},
+  url = {https://brysonkjones.substack.com/p/dissecting-and-open-sourcing-multitask-diffusion-transformer-policy},
+  note = {Blog post}
+}
+```
+
+```bibtex
+@misc{trilbmteam2025carefulexaminationlargebehaviormodels,
+  author       = {TRI LBM Team},
+  title        = {A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulation},
+  year         = {2025},
+  eprint       = {arXiv:2507.05331},
+  archivePrefix = {arXiv},
+  primaryClass = {cs.RO},
+  url          = {https://arxiv.org/abs/2507.05331}
+}
+```
+
+```bibtex
+@misc{bostondynamics2025largebehaviormodelsatlas,
+  author       = {Boston Dynamics and TRI Research Team},
+  title        = {Large Behavior Models and Atlas Find New Footing},
+  year         = {2025},
+  url          = {https://bostondynamics.com/blog/large-behavior-models-atlas-find-new-footing/},
+  note         = {Blog post}
+}
+```

docs/source/reachy2.mdx

@@ -159,6 +159,9 @@ lerobot-record \
     --dataset.fps=15 \
     --dataset.push_to_hub=true \
     --dataset.private=true \
+    --dataset.streaming_encoding=true \
+    --dataset.encoder_threads=2 \
+    # --dataset.vcodec=auto \
     --display_data=true
 ```
 
@@ -198,6 +201,9 @@ lerobot-record \
     --dataset.fps=15 \
     --dataset.push_to_hub=true \
     --dataset.private=true \
+    --dataset.streaming_encoding=true \
+    --dataset.encoder_threads=2 \
+    # --dataset.vcodec=auto \
     --display_data=true
 ```
 

docs/source/rename_map.mdx

@@ -0,0 +1,114 @@
+# Rename Map and Empty Cameras
+
+When you train, evaluate, or record with a robot policy, your **dataset** or **environment** provides observations under one set of keys (e.g. `observation.images.front`, `observation.images.eagle`), while your **policy** expects another (e.g. `observation.images.image`, `observation.images.image2`). The **rename map** bridges that gap without changing the policy or data source.
+
+> **Scope:** The rename map only renames **observation** keys (images and state). Action keys are not affected.
+
+## Why observation keys don't always match
+
+Policies have a fixed set of **input feature names** baked into their pretrained config. For example:
+
+- [pi0fast-libero](https://huggingface.co/lerobot/pi0fast-libero) expects `observation.images.base_0_rgb` and `observation.images.left_wrist_0_rgb`.
+- [xvla-base](https://huggingface.co/lerobot/xvla-base) expects `observation.images.image`, `observation.images.image2`, and `observation.images.image3`.
+
+Your dataset might use different names entirely (e.g. `observation.images.front`, `observation.images.eagle`, `observation.images.glove`), and your eval environment might use yet another set. Rather than editing the policy config or renaming columns in the dataset, you pass a **rename map**: a JSON dictionary that maps source keys to the keys the policy expects. Renaming happens inside the preprocessor pipeline, so the policy always sees its expected keys.
+
+## Using the rename map
+
+Pass the mapping as a JSON string on the command line. The convention is always:
+
+```
+--rename_map='{"source_key": "policy_key", ...}'
+```
+
+where **source_key** is what the dataset or environment provides, and **policy_key** is what the policy expects.
+
+Only listed keys are renamed; everything else passes through unchanged. Order of entries doesn't matter.
+
+Supported policies: **PI0**, **PI05**, **PI0Fast**, **SmolVLA**, and **XVLA**.
+
+### Training
+
+Suppose you fine-tune [lerobot/xvla-base](https://huggingface.co/lerobot/xvla-base) on a dataset with images under `observation.images.front`, `observation.images.eagle`, and `observation.images.glove`. XVLA expects `observation.images.image`, `observation.images.image2`, and `observation.images.image3`:
+
+```bash
+lerobot-train \
+  --dataset.repo_id=YOUR_DATASET \
+  --output_dir=./outputs/xvla_training \
+  --job_name=xvla_training \
+  --policy.path="lerobot/xvla-base" \
+  --policy.repo_id="HF_USER/xvla-your-robot" \
+  --policy.dtype=bfloat16 \
+  --policy.action_mode=auto \
+  --steps=20000 \
+  --policy.device=cuda \
+  --policy.freeze_vision_encoder=false \
+  --policy.freeze_language_encoder=false \
+  --policy.train_policy_transformer=true \
+  --policy.train_soft_prompts=true \
+  --rename_map='{"observation.images.front": "observation.images.image", "observation.images.eagle": "observation.images.image2", "observation.images.glove": "observation.images.image3"}'
+```
+
+### Evaluation
+
+A policy that expects `observation.images.base_0_rgb` and `observation.images.left_wrist_0_rgb` (e.g. [pi0fast-libero](https://huggingface.co/lerobot/pi0fast-libero)), but the LIBERO environment returns `observation.images.image` and `observation.images.image2`:
+
+```bash
+lerobot-eval \
+  --policy.path=lerobot/pi0fast-libero \
+  --env.type=libero \
+  ... \
+  --rename_map='{"observation.images.image": "observation.images.base_0_rgb", "observation.images.image2": "observation.images.left_wrist_0_rgb"}'
+```
+
+### Recording
+
+`lerobot-record` also supports rename maps, nested under the dataset config:
+
+```bash
+lerobot-record \ # When running inference
+  --policy.path="<user>/smolVLA_finetuned" \
+  ... \
+  --dataset.rename_map='{"observation.images.glove2": "observation.images.image"}'
+```
+
+## Alternative: edit the policy config directly
+
+If you always use the same dataset or environment, you can **edit the policy's `config.json`** so its observation keys match your data source. Then no rename map is needed.
+
+The tradeoff: modifying the policy config ties it to one data source. A rename map keeps one policy usable across many datasets and environments.
+
+## Empty cameras: fewer views than the policy expects
+
+Some policies are built for a fixed number of image inputs. If your dataset has fewer cameras, you can set **`empty_cameras`** in the policy config instead of modifying the model architecture.
+
+### How it works
+
+Setting `empty_cameras=N` adds N placeholder image features to the policy config, named:
+
+```
+observation.images.empty_camera_0
+observation.images.empty_camera_1
+...
+```
+
+At runtime, these keys have no corresponding data in the batch. The policy fills them with masked dummy tensors (padded with `-1` for SigLIP-based vision encoders, with a zero attention mask), so the extra image slots are effectively ignored during training and inference.
+
+### Example
+
+XVLA-base has three visual inputs and `empty_cameras=0` by default. Your dataset only has two cameras:
+
+1. Set `--policy.empty_cameras=1`.
+2. The config adds a third key: `observation.images.empty_camera_0`.
+3. Use the rename map for your two real cameras as usual.
+4. The third slot is masked out — no fake images needed in your dataset.
+
+## Quick reference
+
+| Goal                                      | What to do                                                                  |
+| ----------------------------------------- | --------------------------------------------------------------------------- |
+| Dataset keys ≠ policy keys                | `--rename_map='{"dataset_key": "policy_key", ...}'`                         |
+| Env keys ≠ policy keys (eval)             | `--rename_map='{"env_key": "policy_key", ...}'`                             |
+| Recording with different keys (inference) | `--dataset.rename_map='{"source_key": "policy_key", ...}'`.                 |
+| Fewer cameras than policy expects         | `--policy.empty_cameras=N` (supported by PI0, PI05, PI0Fast, SmolVLA, XVLA) |
+| Avoid passing a rename map                | Edit the policy's `config.json` so its keys match your data source          |

docs/source/sarm.mdx

@@ -269,7 +269,7 @@ This generates visualizations showing video frames with subtask boundaries overl
 Train with **no annotations** - uses linear progress from 0 to 1:
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py \
+lerobot-train \
   --dataset.repo_id=your-username/your-dataset \
   --policy.type=sarm \
   --policy.annotation_mode=single_stage \
@@ -288,7 +288,7 @@ python src/lerobot/scripts/lerobot_train.py \
 Train with **dense annotations only** (sparse auto-generated):
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py \
+lerobot-train \
   --dataset.repo_id=your-username/your-dataset \
   --policy.type=sarm \
   --policy.annotation_mode=dense_only \
@@ -307,7 +307,7 @@ python src/lerobot/scripts/lerobot_train.py \
 Train with **both sparse and dense annotations**:
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py \
+lerobot-train \
   --dataset.repo_id=your-username/your-dataset \
   --policy.type=sarm \
   --policy.annotation_mode=dual \
@@ -468,7 +468,7 @@ This script:
 Once you have the progress file, train your policy with RA-BC weighting. The progress file is auto-detected from the dataset path (`sarm_progress.parquet`). Currently PI0, PI0.5 and SmolVLA are supported with RA-BC:
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py \
+lerobot-train \
   --dataset.repo_id=your-username/your-dataset \
   --policy.type=pi0 \
   --use_rabc=true \

docs/source/smolvla.mdx

@@ -106,6 +106,9 @@ lerobot-record \
   --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 \
+  --dataset.streaming_encoding=true \
+  --dataset.encoder_threads=2 \
+  # --dataset.vcodec=auto \
   # <- Teleop optional if you want to teleoperate in between episodes \
   # --teleop.type=so100_leader \
   # --teleop.port=/dev/ttyACM0 \

docs/source/so101.mdx

@@ -236,10 +236,10 @@ It is advisable to install one 3-pin cable in the motor after placing them befor
 
 ### Joint 1
 
+- Install both motor horns. Secure the top horn with a M3x6mm screw. No screws are required for the bottom horn.
 - Place the first motor into the base.
 - Fasten the motor with 4 M2x6mm screws (smallest screws). Two from the top and two from the 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.
@@ -255,9 +255,9 @@ It is advisable to install one 3-pin cable in the motor after placing them befor
 
 ### Joint 2
 
+- Install both motor horns. Secure the top horn with a M3x6mm screw. No screws are required for the bottom horn.
 - 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">
@@ -271,8 +271,8 @@ It is advisable to install one 3-pin cable in the motor after placing them befor
 
 ### 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.
+- Install both motor horns. Secure the top horn with a M3x6mm screw. No screws are required for the bottom horn.
+- Insert motor 3 and fasten using 4 M2x6mm screws.
 - Connect the forearm to motor 3 using 4 M3x6mm screws on each side.
 
 <div class="video-container">
@@ -286,9 +286,10 @@ It is advisable to install one 3-pin cable in the motor after placing them befor
 
 ### Joint 4
 
+- Install both motor horns. Secure the top horn with a M3x6mm screw. No screws are required for the bottom horn.
 - 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.
+- Fasten motor 4 with 4 M2x6mm screws.
 
 <div class="video-container">
   <video controls width="600">
@@ -321,7 +322,7 @@ It is advisable to install one 3-pin cable in the motor after placing them befor
 
 - 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 both motor horns on the gripper motor. Secure the top horn with a M3x6mm screw; no screws are required for the bottom horn.
 - Install the gripper claw and secure it with 4 M3x6mm screws on both sides.
 
 <div class="video-container">

docs/source/streaming_video_encoding.mdx

@@ -0,0 +1,155 @@
+# Streaming Video Encoding Guide
+
+## 1. Overview
+
+Streaming video encoding eliminates the traditional PNG round-trip during video dataset recording. Instead of:
+
+1. Capture frame -> write PNG to disk -> (at episode end) read PNG's -> encode to MP4 -> delete PNG's
+
+Frames can be encoded in real-time during capture:
+
+1. Capture frame -> queue to encoder thread -> encode to MP4 directly
+
+This makes `save_episode()` near-instant (the video is already encoded by the time the episode ends) and removes the blocking wait that previously occurred between episodes, especially with multiple cameras in long episodes.
+
+## 2. Tuning Parameters
+
+| Parameter               | CLI Flag                          | Type          | Default       | Description                                                       |
+| ----------------------- | --------------------------------- | ------------- | ------------- | ----------------------------------------------------------------- |
+| `streaming_encoding`    | `--dataset.streaming_encoding`    | `bool`        | `True`        | Enable real-time encoding during capture                          |
+| `vcodec`                | `--dataset.vcodec`                | `str`         | `"libsvtav1"` | Video codec. `"auto"` detects best HW encoder                     |
+| `encoder_threads`       | `--dataset.encoder_threads`       | `int \| None` | `None` (auto) | Threads per encoder instance. `None` will leave the vcoded decide |
+| `encoder_queue_maxsize` | `--dataset.encoder_queue_maxsize` | `int`         | `60`          | Max buffered frames per camera (~2s at 30fps). Consumes RAM       |
+
+## 3. Performance Considerations
+
+Streaming encoding means the CPU is encoding video **during** the capture loop, not after. This creates a CPU budget that must be shared between:
+
+- **Control loop** (reading cameras, control the robot, writing non-video data)
+- **Encoder threads** (one pool per camera)
+- **Rerun visualization** (if enabled)
+- **OS and other processes**
+
+### Resolution & Number of Cameras Impact
+
+| Setup                     | Throughput (px/sec) | CPU Encoding Load | Notes                          |
+| ------------------------- | ------------------- | ----------------- | ------------------------------ |
+| 2camsx 640x480x3 @30fps   | 55M                 | Low               | Works on most systems          |
+| 2camsx 1280x720x3 @30fps  | 165M                | Moderate          | Comfortable on modern systems  |
+| 2camsx 1920x1080x3 @30fps | 373M                | High              | Requires powerful high-end CPU |
+
+### `encoder_threads` Tuning
+
+This parameter controls how many threads each encoder instance uses internally:
+
+- **Higher values** (e.g., 4-5): Faster encoding, but uses more CPU cores per camera. Good for high-end systems with many cores.
+- **Lower values** (e.g., 1-2): Less CPU per camera, freeing cores for capture and visualization. Good for low-res images and capable CPUs.
+- **`None` (default)**: Lets the codec decide. Information available in the codec logs.
+
+### Backpressure and Frame Dropping
+
+Each camera has a bounded queue (`encoder_queue_maxsize`, default 60 frames). When the encoder can't keep up:
+
+1. The queue fills up (consuming RAM)
+2. New frames are **dropped** (not blocked) — the capture loop continues uninterrupted
+3. A warning is logged: `"Encoder queue full for {camera}, dropped N frame(s)"`
+4. At episode end, total dropped frames per camera are reported
+
+### Symptoms of Encoder Falling Behind
+
+- **System feels laggy and freezes**: all CPUs are at 100%
+- **Dropped frame warnings** in the log or lower frames/FPS than expected in the recorded dataset
+- **Choppy robot movement**: If CPU is severely overloaded, even the capture loop may be affected
+- **Accumulated rerun lag**: Visualization falls behind real-time
+
+## 4. Hardware-Accelerated Encoding
+
+### When to Use
+
+Use HW encoding when:
+
+- CPU is the bottleneck (dropped frames, choppy robot, rerun lag)
+- You have compatible hardware (GPU or dedicated encoder)
+- You're recording at high throughput (high resolution or with many cameras)
+
+### Choosing a Codec
+
+| Codec                 | CPU Usage | File Size      | Quality | Notes                                                            |
+| --------------------- | --------- | -------------- | ------- | ---------------------------------------------------------------- |
+| `libsvtav1` (default) | High      | Smallest       | Best    | Default. Best compression but most CPU-intensive                 |
+| `h264`                | Medium    | ~30-50% larger | Good    | Software H.264. Lower CPU                                        |
+| HW encoders           | Very Low  | Largest        | Good    | Offloads to dedicated hardware. Best for CPU-constrained systems |
+
+### Available HW Encoders
+
+| Encoder             | Platform      | Hardware                                                                                         | CLI Value                            |
+| ------------------- | ------------- | ------------------------------------------------------------------------------------------------ | ------------------------------------ |
+| `h264_videotoolbox` | macOS         | Apple Silicon / Intel                                                                            | `--dataset.vcodec=h264_videotoolbox` |
+| `hevc_videotoolbox` | macOS         | Apple Silicon / Intel                                                                            | `--dataset.vcodec=hevc_videotoolbox` |
+| `h264_nvenc`        | Linux/Windows | NVIDIA GPU                                                                                       | `--dataset.vcodec=h264_nvenc`        |
+| `hevc_nvenc`        | Linux/Windows | NVIDIA GPU                                                                                       | `--dataset.vcodec=hevc_nvenc`        |
+| `h264_vaapi`        | Linux         | Intel/AMD GPU                                                                                    | `--dataset.vcodec=h264_vaapi`        |
+| `h264_qsv`          | Linux/Windows | Intel Quick Sync                                                                                 | `--dataset.vcodec=h264_qsv`          |
+| `auto`              | Any           | Probes the system for available HW encoders. Falls back to `libsvtav1` if no HW encoder is found | `--dataset.vcodec=auto`              |
+
+> [!NOTE]
+> In order to use the HW accelerated encoders you might need to upgrade your GPU drivers.
+
+> [!NOTE]
+> `libsvtav1` is the default because it provides the best training performance; other vcodecs can reduce CPU usage and be faster, but they typically produce larger files and may affect training time.
+
+## 5. Troubleshooting
+
+| Symptom                                                            | Likely Cause                                 | Fix                                                                                                                                                                                                                                                                                  |
+| ------------------------------------------------------------------ | -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| System freezes or choppy robot movement or Rerun visualization lag | CPU starved (100% load usage)                | Close other apps, reduce encoding throughput, lower `encoder_threads`, use `h264`, use `display_data=False`. If the CPU continues to be at 100% then it might be insufficient for your setup, consider `--dataset.streaming_encoding=false` or HW encoding (`--dataset.vcodec=auto`) |
+| "Encoder queue full" warnings or dropped frames in dataset         | Encoder can't keep up (Queue overflow)       | If CPU is not at 100%: Increase `encoder_threads`, increase `encoder_queue_maxsize` or use HW encoding (`--dataset.vcodec=auto`).                                                                                                                                                    |
+| High RAM usage                                                     | Queue filling faster than encoding           | `encoder_threads` too low or CPU insufficient. Reduce `encoder_queue_maxsize` or use HW encoding                                                                                                                                                                                     |
+| Large video files                                                  | Using HW encoder or H.264                    | Expected trade-off. Switch to `libsvtav1` if CPU allows                                                                                                                                                                                                                              |
+| `save_episode()` still slow                                        | `streaming_encoding` is `False`              | Set `--dataset.streaming_encoding=true`                                                                                                                                                                                                                                              |
+| Encoder thread crash                                               | Codec not available or invalid settings      | Check `vcodec` is installed, try `--dataset.vcodec=auto`                                                                                                                                                                                                                             |
+| Recorded dataset is missing frames                                 | CPU/GPU starvation or occasional load spikes | If ~5% of frames are missing, your system is likely overloaded — follow the recommendations above. If fewer frames are missing (~2%), they are probably due to occasional transient load spikes (often at startup) and can be considered expected.                                   |
+
+## 6. Recommended Configurations
+
+These estimates are conservative; we recommend testing them on your setup—start with a low load and increase it gradually.
+
+### High-End Systems: modern 12+ cores (24+ threads)
+
+A throughput between ~250-500M px/sec should be comfortable in CPU. For even better results try HW encoding if available.
+
+```bash
+# 3camsx 1280x720x3 @30fps: Defaults work well. Optionally increase encoder parallelism.
+# 2camsx 1920x1080x3 @30fps: Defaults work well. Optionally increase encoder parallelism.
+lerobot-record --dataset.encoder_threads=5 ...
+
+# 3camsx 1920x1080x3 @30fps: Might require some tuning.
+```
+
+### Mid-Range Systems: modern 8+ cores (16+ threads) or Apple Silicon
+
+A throughput between ~80-300M px/sec should be possible in CPU.
+
+```bash
+# 3camsx 640x480x3 @30fps: Defaults work well. Optionally decrease encoder parallelism.
+# 2camsx 1280x720x3 @30fps: Defaults work well. Optionally decrease encoder parallelism.
+lerobot-record --dataset.encoder_threads=2 ...
+
+# 2camsx 1920x1080x3 @30fps: Might require some tuning.
+```
+
+### Low-Resource Systems: modern 4+ cores (8+ threads) or Raspberry Pi 5
+
+On very constrained systems, streaming encoding may compete too heavily with the capture loop. Disabling it falls back to the PNG-based approach where encoding happens between episodes (blocking, but doesn't interfere with capture). Alternatively, record at a lower throughput to reduce both capture and encoding load. Consider also changing codec to `h264` and using batch encoding.
+
+```bash
+# 2camsx 640x480x3 @30fps: Requires some tuning.
+
+# Use H.264, disable streaming, consider batching encoding
+lerobot-record --dataset.vcodec=h264 --dataset.streaming_encoding=false ...
+```
+
+## 7. Closing note
+
+Performance ultimately depends on your exact setup — frames-per-second, resolution, CPU cores and load, available memory, episode length, and the encoder you choose. Always test with your target workload, be mindful about your CPU & system capabilities and tune `encoder_threads`, `encoder_queue_maxsize`, and
+`vcodec` reasonably. That said, a common practical configuration (for many applications) is three cameras at 640×480x3 @30fps; this usually runs fine with the default streaming video encoding settings in modern systems. Always verify your recorded dataset is healthy by comparing the video duration to the CLI episode duration and confirming the row count equals FPS × CLI duration.

docs/source/unitree_g1.mdx

@@ -1,23 +1,72 @@
 # Unitree G1
 
-This guide covers the complete setup process for the Unitree G1 humanoid, from initial connection to running gr00t_wbc locomotion.
+<img
+  src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/unitree_thumbnail.jpg"
+  alt="Unitree G1 locomanipulation demo"
+  style={{ width: "100%" }}
+/>
 
-## About
-
-We support both 29 and 23 DOF G1 EDU version. We introduce:
-
-- **`unitree g1` robot class, handling low level read/write from/to the humanoid**
-- **ZMQ socket bridge** for remote communication and camera streaming, allowing for remote policy deployment over wlan, eth or directly on the robot
-- **Locomotion policies** from NVIDIA gr00t and Amazon FAR Holosoma
-- **Simulation mode** for testing policies without the physical robot in mujoco
+The Unitree G1 humanoid is now supported in LeRobot! You can teleoperate, train locomanipulation policies, test in sim, and more. Both 29 and 23 DoF variants are supported.
 
 ---
 
-## Connection guide
+## Part 1: Getting Started
 
-### Step 1: Configure Ethernet Interface
+### Install the Unitree SDK
 
-Set a static IP on the same subnet as the robot:
+Follow the [unitree_sdk2_python installation guide](https://github.com/unitreerobotics/unitree_sdk2_python#installation). Tested with `unitree_sdk2py==1.0.1` and `cyclonedds==0.10.2`:
+
+```bash
+conda create -y -n lerobot python=3.12
+conda activate lerobot
+git clone https://github.com/unitreerobotics/unitree_sdk2_python.git
+cd unitree_sdk2_python
+pip install -e .
+cd ..
+```
+
+### Install LeRobot
+
+```bash
+conda install ffmpeg -c conda-forge
+conda install -c conda-forge "pinocchio>=3.0.0,<4.0.0"
+git clone https://github.com/huggingface/lerobot.git
+cd lerobot
+pip install -e '.[unitree_g1]'
+```
+
+<Tip>
+  For now, pinocchio must be installed from conda-forge (not pip) to include the
+  CasADi bindings needed for arm IK.
+</Tip>
+
+### Test the Installation (Simulation)
+
+The simulation environment has its own dependencies. Check the Simulation environment dependencies: [Unitree G1 Mujoco EnvHub](https://huggingface.co/lerobot/unitree-g1-mujoco/tree/main).
+
+```bash
+pip install mujoco loguru msgpack msgpack-numpy
+```
+
+```bash
+lerobot-teleoperate \
+  --robot.type=unitree_g1 \
+  --robot.is_simulation=true \
+  --teleop.type=unitree_g1 \
+  --teleop.id=wbc_unitree \
+  --robot.cameras='{"global_view": {"type": "zmq", "server_address": "localhost", "port": 5555, "camera_name": "head_camera", "width": 640, "height": 480, "fps": 30, "warmup_s": 5}}' \
+  --display_data=true \
+  --robot.controller=GrootLocomotionController
+```
+
+This will launch a [MuJoCo sim instance](https://huggingface.co/lerobot/unitree-g1-mujoco/tree/main) for the G1. You can connect a gamepad to your machine before launching in order to control the robot's locomotion in sim. We support both [HolosomaLocomotionController](https://github.com/amazon-far/holosoma) and [GrootLocomotionController](https://github.com/NVlabs/GR00T-WholeBodyControl) via `--robot.controller`.
+
+- Press `9` to release the robot
+- Press `7` / `8` to increase / decrease waist height
+
+### Connect to the Physical Robot
+
+The G1's Ethernet IP is fixed at `192.168.123.164`. Your machine must have a static IP on the same subnet: `192.168.123.x` where `x ≠ 164`.
 
 ```bash
 # Replace 'enp131s0' with your ethernet interface name (check with `ip a`)
@@ -26,47 +75,23 @@ sudo ip addr add 192.168.123.200/24 dev enp131s0
 sudo ip link set enp131s0 up
 ```
 
-**Note**: The G1's Ethernet IP is fixed at `192.168.123.164`. Your computer must use `192.168.123.x` with x ≠ 164.
-
-### Step 2: SSH into the Robot
+### SSH into the Robot
 
 ```bash
 ssh unitree@192.168.123.164
 # Password: 123
 ```
 
-You should now be connected to the G1's Orin.
+### Share Internet via Ethernet
 
----
-
-## Part 2: Enable WiFi on the Robot
-
-Wlan0 is disabled by default on the G1. To enable it:
-
-### Step 1: Enable WiFi Hardware
-
-```bash
-sudo rfkill unblock wifi
-sudo rfkill unblock all
-
-# Bring up wlan0
-sudo ip link set wlan0 up
-
-# Enable NetworkManager control of wlan0
-sudo nmcli radio wifi on
-sudo nmcli device set wlan0 managed yes
-sudo systemctl restart NetworkManager
-```
-
-### Step 2: Enable Internet Forwarding
+The G1 needs internet access to clone repos and install packages. Share your laptop's connection over Ethernet:
 
 **On your laptop:**
 
 ```bash
-# Enable IP forwarding
 sudo sysctl -w net.ipv4.ip_forward=1
 
-# Set up NAT (replace wlp132s0f0 with your WiFi interface)
+# Replace wlp132s0f0 with your WiFi interface name
 sudo iptables -t nat -A POSTROUTING -o wlp132s0f0 -s 192.168.123.0/24 -j MASQUERADE
 sudo iptables -A FORWARD -i wlp132s0f0 -o enp131s0 -m state --state RELATED,ESTABLISHED -j ACCEPT
 sudo iptables -A FORWARD -i enp131s0 -o wlp132s0f0 -j ACCEPT
@@ -75,217 +100,193 @@ sudo iptables -A FORWARD -i enp131s0 -o wlp132s0f0 -j ACCEPT
 **On the G1:**
 
 ```bash
-# Add laptop as default gateway
 sudo ip route del default 2>/dev/null || true
 sudo ip route add default via 192.168.123.200 dev eth0
 echo "nameserver 8.8.8.8" | sudo tee /etc/resolv.conf
 
-# Test connection
+# Verify
 ping -c 3 8.8.8.8
 ```
 
-### Step 3: Connect to WiFi Network
+### Install the Unitree SDK on the G1
+
+Follow the [unitree_sdk2_python installation guide](https://github.com/unitreerobotics/unitree_sdk2_python#installation):
+
+```bash
+conda create -y -n lerobot python=3.12
+conda activate lerobot
+git clone https://github.com/unitreerobotics/unitree_sdk2_python.git
+cd unitree_sdk2_python
+python -m pip install -e .
+cd ..
+```
+
+### Install LeRobot on the G1
+
+```bash
+git clone https://github.com/huggingface/lerobot.git
+cd lerobot
+conda install -c conda-forge "pinocchio>=3.0.0,<4.0.0"
+python -m pip install -e '.[unitree_g1]'
+```
+
+<Tip>
+  For now, pinocchio must be installed from conda-forge (not pip) to include the
+  CasADi bindings needed for arm IK.
+</Tip>
+
+### (Optional) Enable WiFi on the Robot
+
+For wireless SSH access, you can enable WiFi on the G1 (it's blocked by default):
+
+```bash
+sudo rfkill unblock all
+sudo ip link set wlan0 up
+sudo nmcli radio wifi on
+sudo nmcli device set wlan0 managed yes
+sudo systemctl restart NetworkManager
+```
+
+**Connect to a WiFi network:**
 
 ```bash
-# List available networks
 nmcli device wifi list
 
-# Connect to your WiFi (example)
 sudo nmcli connection add type wifi ifname wlan0 con-name "YourNetwork" ssid "YourNetwork"
 sudo nmcli connection modify "YourNetwork" wifi-sec.key-mgmt wpa-psk
 sudo nmcli connection modify "YourNetwork" wifi-sec.psk "YourPassword"
 sudo nmcli connection modify "YourNetwork" connection.autoconnect yes
 sudo nmcli connection up "YourNetwork"
 
-# Check WiFi IP address
 ip a show wlan0
 ```
 
-### Step 4: SSH Over WiFi
-
-Once connected to WiFi, note the robot's IP address and disconnect the Ethernet cable. You can now SSH over WiFi:
+You can then SSH over WiFi instead of Ethernet:
 
 ```bash
-ssh unitree@<YOUR_ROBOT_IP>
+ssh unitree@<ROBOT_WIFI_IP>
 # Password: 123
 ```
 
-Replace `<YOUR_ROBOT_IP>` with your robot's actual WiFi IP address.
+---
+
+## Part 2: Teleoperation & Locomotion
+
+### Run the Robot Server
+
+On the robot (from `~/lerobot`):
+
+```bash
+cd ~/lerobot
+python src/lerobot/robots/unitree_g1/run_g1_server.py --camera
+```
+
+### Run the Locomotion Policy
+
+You can run the teleoperation client from your laptop over Ethernet, over WiFi (experimental), or directly on the robot itself. Mind potential latency introduced by your network.
+
+**From your laptop:**
+
+```bash
+lerobot-teleoperate \
+  --robot.type=unitree_g1 \
+  --robot.is_simulation=false \
+  --robot.robot_ip=<ROBOT_IP> \
+  --teleop.type=unitree_g1 \
+  --teleop.id=wbc_unitree \
+  --robot.cameras='{"global_view": {"type": "zmq", "server_address": "<ROBOT_IP>", "port": 5555, "camera_name": "head_camera", "width": 640, "height": 480, "fps": 30}}' \
+  --display_data=true \
+  --robot.controller=HolosomaLocomotionController
+```
+
+We support both [GrootLocomotionController](https://github.com/NVlabs/GR00T-WholeBodyControl) and [HolosomaLocomotionController](https://github.com/amazon-far/holosoma) via `--robot.controller`.
 
 ---
 
-## Part 3: Robot Server Setup
+## Part 3: Loco-Manipulation with the Homunculus Exoskeleton
 
-### Step 1: Install LeRobot on the Orin
+We provide a loco-manipulation solution via the Homunculus Exoskeleton — an open-source 7 DoF exoskeleton for whole-body control. Check it out [here](https://github.com/nepyope/hmc_exo).
 
-SSH into the robot and install LeRobot:
-
-```bash
-ssh unitree@<YOUR_ROBOT_IP>
-
-conda create -y -n lerobot python=3.10
-conda activate lerobot
-git clone https://github.com/huggingface/lerobot.git
-cd lerobot
-pip install -e '.[unitree_g1]'
-git clone https://github.com/unitreerobotics/unitree_sdk2_python.git
-cd unitree_sdk2_python  && pip install -e .
-```
-
-**Note**: The Unitree SDK requires CycloneDDS v0.10.2 to be installed. See the [Unitree SDK documentation](https://github.com/unitreerobotics/unitree_sdk2_python) for details.
-
-### Step 2: Run the Robot Server
-
-On the robot:
-
-```bash
-python src/lerobot/robots/unitree_g1/run_g1_server.py
-```
-
-**Important**: Keep this terminal running. The server must be active for remote control.
-
----
-
-## Part 4: Controlling the robot
-
-With the robot server running, you can now control the robot remotely. Let's launch a locomotion policy
-
-### Step 1: Install LeRobot on your machine
-
-```bash
-conda create -y -n lerobot python=3.10
-conda activate lerobot
-git clone https://github.com/huggingface/lerobot.git
-cd lerobot
-pip install -e '.[unitree_g1]'
-git clone https://github.com/unitreerobotics/unitree_sdk2_python.git
-cd unitree_sdk2_python  && pip install -e .
-```
-
-### Step 2: Update Robot IP in Config
-
-Edit the config file to match your robot's WiFi IP:
-
-```python
-# In src/lerobot/robots/unitree_g1/config_unitree_g1.py
-robot_ip: str = "<YOUR_ROBOT_IP>"  # Replace with your robot's WiFi IP.
-```
-
-### Step 3: Run the Locomotion Policy
-
-```bash
-# Run GR00T locomotion controller
-python examples/unitree_g1/gr00t_locomotion.py --repo-id "nepyope/GR00T-WholeBodyControl_g1"
-
-# Run Holosoma locomotion controller
-python examples/unitree_g1/holosoma_locomotion.py
-
-```
-
-Press `Ctrl+C` to stop the policy.
-
----
-
-## Running in Simulation Mode (MuJoCo)
-
-You can test policies before deploying on the physical robot using MuJoCo simulation. Set `is_simulation=True` in config or pass `--robot.is_simulation=true` via CLI.
-
-### Calibrate Exoskeleton Teleoperator
+### Calibrate
 
 ```bash
 lerobot-calibrate \
-    --teleop.type=unitree_g1 \
-    --teleop.left_arm_config.port=/dev/ttyACM1 \
-    --teleop.right_arm_config.port=/dev/ttyACM0 \
-    --teleop.id=exo
+  --teleop.type=unitree_g1 \
+  --teleop.left_arm_config.port=/dev/ttyACM1 \
+  --teleop.right_arm_config.port=/dev/ttyACM0 \
+  --teleop.id=exo
 ```
 
-### Teleoperate in Simulation
+During calibration move each joint through its entire range. After fitting, move the joint in a neutral position and press `n` to advance.
+
+### Record a Dataset
 
 ```bash
-lerobot-teleoperate \
-    --robot.type=unitree_g1 \
-    --robot.is_simulation=true \
-    --teleop.type=unitree_g1 \
-    --teleop.left_arm_config.port=/dev/ttyACM1 \
-    --teleop.right_arm_config.port=/dev/ttyACM0 \
-    --teleop.id=exo \
-    --fps=100
+lerobot-record \
+  --robot.type=unitree_g1 \
+  --robot.is_simulation=true \
+  --robot.cameras='{"global_view": {"type": "zmq", "server_address": "localhost", "port": 5555, "camera_name": "head_camera", "width": 640, "height": 480, "fps": 30}}' \
+  --teleop.type=unitree_g1 \
+  --teleop.left_arm_config.port=/dev/ttyACM1 \
+  --teleop.right_arm_config.port=/dev/ttyACM0 \
+  --teleop.id=exo \
+  --dataset.repo_id=your-username/dataset-name \
+  --dataset.single_task="Test" \
+  --dataset.num_episodes=2 \
+  --dataset.episode_time_s=5 \
+  --dataset.reset_time_s=5 \
+  --dataset.push_to_hub=true \
+  --dataset.streaming_encoding=true \
+  --dataset.encoder_threads=2
 ```
 
-### Record Dataset in Simulation
+> **Note:** Omit `--teleop.left_arm_config.port` and `--teleop.right_arm_config.port` if you're only using the joystick.
 
-```bash
-python -m lerobot.scripts.lerobot_record \
-    --robot.type=unitree_g1 \
-    --robot.is_simulation=true \
-    --robot.cameras='{"global_view": {"type": "zmq", "server_address": "localhost", "port": 5555, "camera_name": "head_camera", "width": 640, "height": 480, "fps": 30}}' \
-    --teleop.type=unitree_g1 \
-    --teleop.left_arm_config.port=/dev/ttyACM1 \
-    --teleop.right_arm_config.port=/dev/ttyACM0 \
-    --teleop.id=exo \
-    --dataset.repo_id=your-username/dataset-name \
-    --dataset.single_task="Test" \
-    --dataset.num_episodes=2 \
-    --dataset.episode_time_s=5 \
-    --dataset.reset_time_s=5 \
-    --dataset.push_to_hub=true
-```
-
-Example simulation dataset: [nepyope/teleop_test_sim](https://huggingface.co/datasets/nepyope/teleop_test_sim)
+Example dataset: [nepyope/unitree_box_move_blue_full](https://huggingface.co/datasets/nepyope/unitree_box_move_blue_full)
 
 ---
 
-## Running on Real Robot
+## Part 4: Training & Inference
 
-Once the robot server is running on the G1 (see Part 3), you can teleoperate and record on the real robot.
-
-### Start the Camera Server
-
-On the robot, start the ZMQ image server:
+### Train
 
 ```bash
-python src/lerobot/cameras/zmq/image_server.py
+python src/lerobot/scripts/lerobot_train.py \
+  --dataset.repo_id=your-username/dataset-name  \
+  --policy.type=pi05 \
+  --output_dir=./outputs/pi05_training \
+  --job_name=pi05_training \
+  --policy.repo_id=your-username/your-repo-id \
+  --policy.pretrained_path=lerobot/pi05_base \
+  --policy.compile_model=true \
+  --policy.gradient_checkpointing=true \
+  --wandb.enable=true \
+  --policy.dtype=bfloat16 \
+  --policy.freeze_vision_encoder=false \
+  --policy.train_expert_only=false \
+  --steps=3000 \
+  --policy.device=cuda \
+  --batch_size=32
 ```
 
-Keep this running in a separate terminal for camera streaming during recording.
+### Inference with RTC
 
-### Teleoperate Real Robot
+Once trained, we recommend deploying policies using inference-time RTC:
 
 ```bash
-lerobot-teleoperate \
-    --robot.type=unitree_g1 \
-    --robot.is_simulation=false \
-    --teleop.type=unitree_g1 \
-    --teleop.left_arm_config.port=/dev/ttyACM1 \
-    --teleop.right_arm_config.port=/dev/ttyACM0 \
-    --teleop.id=exo \
-    --fps=100
+python examples/rtc/eval_with_real_robot.py \
+  --policy.path=your-username/your-repo-id \
+  --policy.device=cuda \
+  --robot.type=unitree_g1 \
+  --robot.is_simulation=false \
+  --robot.controller=HolosomaLocomotionController \
+  --robot.cameras='{"global_view": {"type": "zmq", "server_address": "<ROBOT_IP>", "port": 5555, "camera_name": "head_camera", "width": 640, "height": 480, "fps": 30}}' \
+  --task="task_description" \
+  --duration=1000 \
+  --fps=30 \
+  --rtc.enabled=true
 ```
 
-### Record Dataset on Real Robot
-
-```bash
-python -m lerobot.scripts.lerobot_record \
-    --robot.type=unitree_g1 \
-    --robot.is_simulation=false \
-    --robot.cameras='{"global_view": {"type": "zmq", "server_address": "172.18.129.215", "port": 5555, "camera_name": "head_camera", "width": 640, "height": 480, "fps": 30}}' \
-    --teleop.type=unitree_g1 \
-    --teleop.left_arm_config.port=/dev/ttyACM1 \
-    --teleop.right_arm_config.port=/dev/ttyACM0 \
-    --teleop.id=exo \
-    --dataset.repo_id=your-username/dataset-name \
-    --dataset.single_task="Test" \
-    --dataset.num_episodes=2 \
-    --dataset.episode_time_s=5 \
-    --dataset.reset_time_s=5 \
-    --dataset.push_to_hub=true
-```
-
-**Note**: Update `server_address` to match your robot's camera server IP.
-
-Example real robot dataset: [nepyope/teleop_test_real](https://huggingface.co/datasets/nepyope/teleop_test_real)
-
 ---
 
 ## Additional Resources
@@ -294,8 +295,8 @@ Example real robot dataset: [nepyope/teleop_test_real](https://huggingface.co/da
 - [GR00T-WholeBodyControl](https://github.com/NVlabs/GR00T-WholeBodyControl)
 - [Holosoma](https://github.com/amazon-far/holosoma)
 - [LeRobot Documentation](https://github.com/huggingface/lerobot)
-- [Unitree_IL_Lerobot](https://github.com/unitreerobotics/unitree_IL_lerobot)
+- [Unitree IL LeRobot](https://github.com/unitreerobotics/unitree_IL_lerobot)
 
 ---
 
-_Last updated: December 2025_
+_Last updated: March 2026_

docs/source/using_dataset_tools.mdx

@@ -12,6 +12,7 @@ LeRobot provides several utilities for manipulating datasets:
 4. **Add Features** - Add new features to a dataset
 5. **Remove Features** - Remove features from a dataset
 6. **Convert to Video** - Convert image-based datasets to video format for efficient storage
+7. **Show the Info of Datasets** - Show the summary of datasets information such as number of episode etc.
 
 The core implementation is in `lerobot.datasets.dataset_tools`.
 An example script detailing how to use the tools API is available in `examples/dataset/use_dataset_tools.py`.
@@ -156,6 +157,30 @@ lerobot-edit-dataset \
 
 **Note:** The resulting dataset will be a proper LeRobotDataset with all cameras encoded as videos in the `videos/` directory, with parquet files containing only metadata (no raw image data). All episodes, stats, and tasks are preserved.
 
+### Show the information of datasets
+
+Show the information of datasets such as number of episode, number of frame, File size and so on.
+No change will be made to the dataset
+
+```bash
+
+# Show dataset information without feature details
+lerobot-edit-dataset \
+    --repo_id lerobot/pusht_image \
+    --operation.type info \
+
+# Show dataset information with feature details
+lerobot-edit-dataset \
+    --repo_id lerobot/pusht_image \
+    --operation.type info \
+    --operation.show_features true
+
+```
+
+**Parameters:**
+
+- `parameters`: The flag to control show or no show dataset information with feature details.(default=false)
+
 ### Push to Hub
 
 Add the `--push_to_hub true` flag to any command to automatically upload the resulting dataset to the Hugging Face Hub:

docs/source/walloss.mdx

@@ -45,7 +45,7 @@ policy.type=wall_x
 For training WallX, you can use the standard LeRobot training script with the appropriate configuration:
 
 ```bash
-python src/lerobot/scripts/lerobot_train.py \
+lerobot-train \
     --dataset.repo_id=your_dataset \
     --policy.type=wall_x \
     --output_dir=./outputs/wallx_training \

docs/source/xvla.mdx

@@ -154,7 +154,7 @@ lerobot-train \
 
 ```bash
 lerobot-train \
-  --dataset.repo_id=pepijn223/bimanual-so100-handover-cube \
+  --dataset.repo_id=<USER>/bimanual-so100-handover-cube \
   --output_dir=./outputs/xvla_bimanual \
   --job_name=xvla_so101_training \
   --policy.path="lerobot/xvla-base" \

examples/backward_compatibility/replay.py

@@ -22,7 +22,7 @@ lerobot-replay \
     --robot.type=so100_follower \
     --robot.port=/dev/tty.usbmodem58760431541 \
     --robot.id=black \
-    --dataset.repo_id=aliberts/record-test \
+    --dataset.repo_id=<USER>/record-test \
     --dataset.episode=2
 ```
 """
@@ -57,7 +57,7 @@ class DatasetReplayConfig:
     repo_id: str
     # Episode to replay.
     episode: int
-    # Root directory where the dataset will be stored (e.g. 'dataset/path').
+    # Root directory where the dataset will be stored (e.g. 'dataset/path'). If None, defaults to $HF_LEROBOT_HOME/repo_id.
     root: str | Path | None = None
     # Limit the frames per second. By default, uses the policy fps.
     fps: int = 30
@@ -78,7 +78,7 @@ def replay(cfg: ReplayConfig):
 
     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)
+    actions = dataset.select_columns(ACTION)
     robot.connect()
 
     try:

examples/dataset/load_lerobot_dataset.py

@@ -32,7 +32,8 @@ import torch
 from huggingface_hub import HfApi
 
 import lerobot
-from lerobot.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 
 
 def main():
@@ -87,9 +88,8 @@ def main():
     # The previous metadata class is contained in the 'meta' attribute of the dataset:
     print(dataset.meta)
 
-    # LeRobotDataset actually wraps an underlying Hugging Face dataset
-    # (see https://huggingface.co/docs/datasets for more information).
-    print(dataset.hf_dataset)
+    # You can inspect the dataset using its repr:
+    print(dataset)
 
     # LeRobot datasets also subclasses PyTorch datasets so you can do everything you know and love from working
     # with the latter, like iterating through the dataset.

examples/dataset/slurm_compute_rabc.py

@@ -0,0 +1,490 @@
+#!/usr/bin/env python
+
+# Copyright 2025 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.
+
+"""
+SLURM-distributed SARM RA-BC annotation pipeline.
+
+Computes SARM progress values for all frames in a dataset, distributed across
+SLURM workers, then merges the shards into a single sarm_progress.parquet.
+
+Two subcommands, each a separate SLURM submission:
+
+  compute    – N workers, each computes progress for a subset of episodes
+  aggregate  – 1 worker, merges N shards into sarm_progress.parquet, pushes to hub
+
+Usage:
+    python slurm_compute_rabc.py compute \\
+        --repo-id user/dataset --reward-model-path user/sarm_model \\
+        --stride 10 --device cpu --workers 50 --partition cpu
+
+    python slurm_compute_rabc.py aggregate \\
+        --repo-id user/dataset --reward-model-path user/sarm_model \\
+        --partition cpu --push-to-hub
+"""
+
+import argparse
+from pathlib import Path
+
+from datatrove.executor import LocalPipelineExecutor
+from datatrove.executor.slurm import SlurmPipelineExecutor
+from datatrove.pipeline.base import PipelineStep
+
+
+class ComputeProgressShards(PipelineStep):
+    """Each worker computes SARM progress for its assigned episodes."""
+
+    def __init__(
+        self, repo_id, reward_model_path, stride=1, head_mode="sparse", device="cpu", shard_dir="rabc_shards"
+    ):
+        super().__init__()
+        if stride < 1:
+            raise ValueError(f"stride must be >= 1, got {stride}")
+        self.repo_id = repo_id
+        self.reward_model_path = reward_model_path
+        self.stride = stride
+        self.head_mode = head_mode
+        self.device = device
+        self.shard_dir = shard_dir
+
+    def run(self, data=None, rank: int = 0, world_size: int = 1):
+        import logging
+        from pathlib import Path
+
+        import numpy as np
+        import pyarrow as pa
+        import pyarrow.parquet as pq
+        import torch
+        from tqdm import tqdm
+
+        from lerobot.policies.sarm.compute_rabc_weights import (
+            generate_all_frame_indices,
+            interpolate_progress,
+            load_sarm_resources,
+        )
+        from lerobot.utils.utils import init_logging
+
+        init_logging()
+
+        dataset, reward_model, preprocess = load_sarm_resources(
+            self.repo_id,
+            self.reward_model_path,
+            self.device,
+        )
+
+        if hasattr(preprocess, "eval"):
+            preprocess.eval()
+        for step in preprocess.steps:
+            if hasattr(step, "eval"):
+                step.eval()
+
+        image_key = reward_model.config.image_key
+        state_key = reward_model.config.state_key
+        frame_gap = reward_model.config.frame_gap
+        center_idx = reward_model.config.n_obs_steps // 2
+
+        dual_mode = reward_model.config.uses_dual_heads
+        compute_sparse = self.head_mode in ("sparse", "both") or not dual_mode
+        compute_dense = self.head_mode in ("dense", "both") and dual_mode
+
+        my_episodes = list(range(dataset.num_episodes))[rank::world_size]
+        if not my_episodes:
+            logging.info(f"Rank {rank}: no episodes assigned")
+            return
+        logging.info(f"Rank {rank}: {len(my_episodes)} / {dataset.num_episodes} episodes")
+
+        all_rows = []
+
+        for ep_idx in tqdm(my_episodes, desc=f"Rank {rank}"):
+            ep = dataset.meta.episodes[ep_idx]
+            ep_start, ep_end = ep["dataset_from_index"], ep["dataset_to_index"]
+            task = dataset[ep_start].get("task", "perform the task")
+
+            all_ep_indices = generate_all_frame_indices(ep_start, ep_end, frame_gap)
+            if self.stride > 1:
+                compute_indices = [i for i in all_ep_indices if (i - ep_start) % self.stride == 0]
+                if (ep_end - 1) not in compute_indices:
+                    compute_indices.append(ep_end - 1)
+                compute_indices = sorted(set(compute_indices))
+            else:
+                compute_indices = all_ep_indices
+
+            frame_results = {}
+            for qi in tqdm(compute_indices, desc=f"  Ep {ep_idx}", leave=False):
+                try:
+                    sample = dataset[qi]
+                    batch = {
+                        image_key: sample[image_key],
+                        "task": task,
+                        "index": qi,
+                        "episode_index": ep_idx,
+                    }
+                    if state_key in sample:
+                        batch[state_key] = sample[state_key]
+
+                    with torch.no_grad():
+                        processed = preprocess(batch)
+                        vf = processed["video_features"].to(self.device)
+                        tf = processed["text_features"].to(self.device)
+                        sf = processed.get("state_features")
+                        if sf is not None:
+                            sf = sf.to(self.device)
+                        lengths = processed.get("lengths")
+
+                        sparse_val = dense_val = np.nan
+                        if compute_sparse:
+                            r = reward_model.calculate_rewards(
+                                text_embeddings=tf,
+                                video_embeddings=vf,
+                                state_features=sf,
+                                lengths=lengths,
+                                return_all_frames=True,
+                                head_mode="sparse",
+                            )
+                            sparse_val = float(r[0, center_idx] if r.ndim == 2 else r[center_idx])
+                        if compute_dense:
+                            r = reward_model.calculate_rewards(
+                                text_embeddings=tf,
+                                video_embeddings=vf,
+                                state_features=sf,
+                                lengths=lengths,
+                                return_all_frames=True,
+                                head_mode="dense",
+                            )
+                            dense_val = float(r[0, center_idx] if r.ndim == 2 else r[center_idx])
+
+                        frame_results[qi] = (sparse_val, dense_val)
+                except Exception as e:
+                    logging.warning(f"Failed frame {qi}: {e}")
+
+            if not frame_results:
+                logging.warning(f"Episode {ep_idx}: all frames failed, skipping")
+                continue
+
+            # Interpolate to all frames in this episode
+            computed_idx = np.array(sorted(frame_results.keys()))
+            all_frame_arr = np.arange(ep_start, ep_end)
+
+            sparse_vals = np.array([frame_results[i][0] for i in computed_idx]) if compute_sparse else None
+            dense_vals = np.array([frame_results[i][1] for i in computed_idx]) if compute_dense else None
+
+            if self.stride > 1 and len(computed_idx) > 1:
+                if compute_sparse:
+                    sparse_vals = interpolate_progress(computed_idx, sparse_vals, all_frame_arr)
+                if compute_dense:
+                    dense_vals = interpolate_progress(computed_idx, dense_vals, all_frame_arr)
+                output_frames = all_frame_arr
+            else:
+                # Use only successfully computed frames to avoid indexing mismatch on failures
+                output_frames = computed_idx
+
+            for i, fi in enumerate(output_frames):
+                row = {"index": int(fi), "episode_index": ep_idx, "frame_index": int(fi - ep_start)}
+                if compute_sparse:
+                    row["progress_sparse"] = float(sparse_vals[i])
+                if compute_dense:
+                    row["progress_dense"] = float(dense_vals[i])
+                all_rows.append(row)
+
+        if all_rows:
+            import pandas as pd
+
+            df = pd.DataFrame(all_rows).sort_values("index").reset_index(drop=True)
+            table = pa.Table.from_pandas(df, preserve_index=False)
+            table = table.replace_schema_metadata({b"reward_model_path": self.reward_model_path.encode()})
+            shard_dir = Path(self.shard_dir)
+            shard_dir.mkdir(parents=True, exist_ok=True)
+            out = shard_dir / f"shard_{rank:05d}.parquet"
+            pq.write_table(table, out)
+            logging.info(f"Rank {rank}: saved {len(df)} rows to {out}")
+
+
+class AggregateProgress(PipelineStep):
+    """Merge all shard parquets into final sarm_progress.parquet."""
+
+    def __init__(self, repo_id, reward_model_path, shard_dir="rabc_shards", push_to_hub=False):
+        super().__init__()
+        self.repo_id = repo_id
+        self.reward_model_path = reward_model_path
+        self.shard_dir = shard_dir
+        self.push_to_hub = push_to_hub
+
+    def run(self, data=None, rank: int = 0, world_size: int = 1):
+        import datetime
+        import logging
+        import os
+        from pathlib import Path
+
+        import pandas as pd
+        import pyarrow as pa
+        import pyarrow.parquet as pq
+
+        from lerobot.datasets.lerobot_dataset import LeRobotDataset
+        from lerobot.utils.utils import init_logging
+
+        init_logging()
+        if rank != 0:
+            return
+
+        shard_dir = Path(self.shard_dir)
+        shards = sorted(shard_dir.glob("shard_*.parquet"))
+        if not shards:
+            raise FileNotFoundError(f"No shards found in {shard_dir}")
+
+        # Log shard modification time range to help detect stale files
+        mtimes = [os.path.getmtime(s) for s in shards]
+        oldest = datetime.datetime.fromtimestamp(min(mtimes)).isoformat(timespec="seconds")
+        newest = datetime.datetime.fromtimestamp(max(mtimes)).isoformat(timespec="seconds")
+        logging.info(f"Aggregating {len(shards)} shards (oldest: {oldest}, newest: {newest})")
+
+        df = pd.concat([pd.read_parquet(s) for s in shards], ignore_index=True)
+        df = df.sort_values("index").reset_index(drop=True)
+
+        table = pa.Table.from_pandas(df, preserve_index=False)
+        table = table.replace_schema_metadata({b"reward_model_path": self.reward_model_path.encode()})
+
+        temp_ds = LeRobotDataset(self.repo_id, download_videos=False)
+        out_path = Path(temp_ds.root) / "sarm_progress.parquet"
+        out_path.parent.mkdir(parents=True, exist_ok=True)
+        pq.write_table(table, out_path)
+        logging.info(f"Saved {len(df)} rows to {out_path}")
+
+        for col in ["progress_sparse", "progress_dense"]:
+            if col in df.columns:
+                v = df[col].dropna()
+                logging.info(
+                    f"{col}: mean={v.mean():.4f} std={v.std():.4f} min={v.min():.4f} max={v.max():.4f}"
+                )
+
+        if self.push_to_hub:
+            from huggingface_hub import HfApi
+
+            api = HfApi()
+            hub_path = "sarm_progress.parquet"
+            logging.info(f"Uploading to {self.repo_id}/{hub_path}")
+            api.upload_file(
+                path_or_fileobj=str(out_path),
+                path_in_repo=hub_path,
+                repo_id=self.repo_id,
+                repo_type="dataset",
+            )
+            logging.info(f"Uploaded: https://huggingface.co/datasets/{self.repo_id}/blob/main/{hub_path}")
+
+
+def make_compute_executor(
+    repo_id,
+    reward_model_path,
+    stride,
+    head_mode,
+    device,
+    shard_dir,
+    logs_dir,
+    job_name,
+    slurm,
+    workers,
+    partition,
+    cpus_per_task,
+    mem_per_cpu,
+):
+    kwargs = {
+        "pipeline": [
+            ComputeProgressShards(repo_id, reward_model_path, stride, head_mode, device, str(shard_dir)),
+        ],
+        "logging_dir": str(logs_dir / job_name),
+    }
+
+    if slurm:
+        kwargs.update(
+            {
+                "job_name": job_name,
+                "tasks": workers,
+                "workers": workers,
+                "time": "24:00:00",
+                "partition": partition,
+                "cpus_per_task": cpus_per_task,
+                "sbatch_args": {"mem-per-cpu": mem_per_cpu},
+            }
+        )
+        return SlurmPipelineExecutor(**kwargs)
+
+    kwargs.update({"tasks": workers, "workers": 1})
+    return LocalPipelineExecutor(**kwargs)
+
+
+def make_aggregate_executor(
+    repo_id,
+    reward_model_path,
+    shard_dir,
+    logs_dir,
+    job_name,
+    slurm,
+    partition,
+    cpus_per_task,
+    mem_per_cpu,
+    push_to_hub,
+):
+    kwargs = {
+        "pipeline": [
+            AggregateProgress(repo_id, reward_model_path, str(shard_dir), push_to_hub),
+        ],
+        "logging_dir": str(logs_dir / job_name),
+    }
+
+    if slurm:
+        kwargs.update(
+            {
+                "job_name": job_name,
+                "tasks": 1,
+                "workers": 1,
+                "time": "02:00:00",
+                "partition": partition,
+                "cpus_per_task": cpus_per_task,
+                "sbatch_args": {"mem-per-cpu": mem_per_cpu},
+            }
+        )
+        return SlurmPipelineExecutor(**kwargs)
+
+    kwargs.update({"tasks": 1, "workers": 1})
+    return LocalPipelineExecutor(**kwargs)
+
+
+def _add_shared_args(p):
+    p.add_argument(
+        "--repo-id",
+        type=str,
+        required=True,
+        help="Hugging Face repository identifier, e.g. 'user/dataset'.",
+    )
+    p.add_argument(
+        "--shard-dir",
+        type=Path,
+        default=Path("rabc_shards"),
+        help="Directory to read/write per-rank parquet shards.",
+    )
+    p.add_argument(
+        "--logs-dir",
+        type=Path,
+        default=Path("logs"),
+        help="Directory for datatrove logs.",
+    )
+    p.add_argument(
+        "--job-name",
+        type=str,
+        default=None,
+        help="SLURM job name (defaults to rabc_<subcommand>).",
+    )
+    p.add_argument(
+        "--slurm",
+        type=int,
+        default=1,
+        help="1 = submit via SLURM; 0 = run locally (useful for debugging).",
+    )
+    p.add_argument(
+        "--partition",
+        type=str,
+        default=None,
+        help="SLURM partition to submit to.",
+    )
+    p.add_argument(
+        "--cpus-per-task",
+        type=int,
+        default=4,
+        help="Number of CPUs per SLURM task.",
+    )
+    p.add_argument(
+        "--mem-per-cpu",
+        type=str,
+        default="4G",
+        help="Memory per CPU, e.g. '4G' or '1950M'.",
+    )
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="SLURM-distributed SARM RA-BC annotation pipeline",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    sub = parser.add_subparsers(dest="command", required=True)
+
+    # compute subcommand
+    cp = sub.add_parser(
+        "compute",
+        help="Distribute progress computation across SLURM workers.",
+    )
+    _add_shared_args(cp)
+    cp.add_argument(
+        "--reward-model-path",
+        type=str,
+        required=True,
+        help="Path or HF repo id of the SARM reward model.",
+    )
+    cp.add_argument(
+        "--stride",
+        type=int,
+        default=1,
+        help="Compute every Nth frame; intermediate frames are interpolated (must be >= 1).",
+    )
+    cp.add_argument(
+        "--head-mode",
+        type=str,
+        default="sparse",
+        choices=["sparse", "dense", "both"],
+        help="Which reward head(s) to compute.",
+    )
+    cp.add_argument(
+        "--device",
+        type=str,
+        default="cpu",
+        help="Device for reward model inference, e.g. 'cpu' or 'cuda'.",
+    )
+    cp.add_argument(
+        "--workers",
+        type=int,
+        default=50,
+        help="Number of parallel SLURM tasks (one shard per worker).",
+    )
+
+    # aggregate subcommand
+    ap = sub.add_parser(
+        "aggregate",
+        help="Merge per-rank shards into a single sarm_progress.parquet.",
+    )
+    _add_shared_args(ap)
+    ap.add_argument(
+        "--reward-model-path",
+        type=str,
+        required=True,
+        help="Path or HF repo id of the SARM reward model (stored in parquet metadata).",
+    )
+    ap.add_argument(
+        "--push-to-hub",
+        action="store_true",
+        help="Upload sarm_progress.parquet to the Hugging Face Hub after aggregation.",
+    )
+
+    args = parser.parse_args()
+    job_name = args.job_name or f"rabc_{args.command}"
+    kwargs = vars(args)
+    kwargs["slurm"] = kwargs.pop("slurm") == 1
+    kwargs["job_name"] = job_name
+    command = kwargs.pop("command")
+
+    executor = make_compute_executor(**kwargs) if command == "compute" else make_aggregate_executor(**kwargs)
+
+    executor.run()
+
+
+if __name__ == "__main__":
+    main()

examples/lekiwi/evaluate.py

@@ -14,8 +14,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from lerobot.datasets.feature_utils import hw_to_dataset_features
 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.policies.factory import make_pre_post_processors
 from lerobot.processor import make_default_processors

examples/lekiwi/record.py

@@ -14,8 +14,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from lerobot.datasets.feature_utils import hw_to_dataset_features
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
-from lerobot.datasets.utils import hw_to_dataset_features
 from lerobot.processor import make_default_processors
 from lerobot.robots.lekiwi.config_lekiwi import LeKiwiClientConfig
 from lerobot.robots.lekiwi.lekiwi_client import LeKiwiClient

examples/lekiwi/replay.py

@@ -35,9 +35,7 @@ def main():
 
     # Fetch the dataset to replay
     dataset = LeRobotDataset("<hf_username>/<dataset_repo_id>", episodes=[EPISODE_IDX])
-    # Filter dataset to only include frames from the specified episode since episodes are chunked in dataset V3.0
-    episode_frames = dataset.hf_dataset.filter(lambda x: x["episode_index"] == EPISODE_IDX)
-    actions = episode_frames.select_columns(ACTION)
+    actions = dataset.select_columns(ACTION)
 
     # Connect to the robot
     robot.connect()
@@ -48,7 +46,7 @@ def main():
 
         print("Starting replay loop...")
         log_say(f"Replaying episode {EPISODE_IDX}")
-        for idx in range(len(episode_frames)):
+        for idx in range(dataset.num_frames):
             t0 = time.perf_counter()
 
             # Get recorded action from dataset

examples/lekiwi/teleoperate.py

@@ -43,12 +43,13 @@ def main():
     keyboard.connect()
 
     # Init rerun viewer
-    init_rerun(session_name="lekiwi_teleop")
+    init_rerun(session_name="lekiwi_teleop", robot=robot, reset_time=True)
 
     if not robot.is_connected or not leader_arm.is_connected or not keyboard.is_connected:
         raise ValueError("Robot or teleop is not connected!")
 
     print("Starting teleop loop...")
+    start = time.perf_counter()
     while True:
         t0 = time.perf_counter()
 
@@ -69,7 +70,7 @@ def main():
         _ = robot.send_action(action)
 
         # Visualize
-        log_rerun_data(observation=observation, action=action)
+        log_rerun_data(observation=observation, action=action, log_time=time.perf_counter() - start)
 
         precise_sleep(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
 

examples/phone_to_so100/evaluate.py

@@ -16,15 +16,13 @@
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
 from lerobot.configs.types import FeatureType, PolicyFeature
+from lerobot.datasets.feature_utils import combine_feature_dicts
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.datasets.pipeline_features import aggregate_pipeline_dataset_features, create_initial_features
-from lerobot.datasets.utils import combine_feature_dicts
 from lerobot.model.kinematics import RobotKinematics
 from lerobot.policies.act.modeling_act import ACTPolicy
 from lerobot.policies.factory import make_pre_post_processors
 from lerobot.processor import (
-    RobotAction,
-    RobotObservation,
     RobotProcessorPipeline,
     make_default_teleop_action_processor,
 )
@@ -40,6 +38,7 @@ from lerobot.robots.so_follower.robot_kinematic_processor import (
     InverseKinematicsEEToJoints,
 )
 from lerobot.scripts.lerobot_record import record_loop
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.control_utils import init_keyboard_listener
 from lerobot.utils.utils import log_say
 from lerobot.utils.visualization_utils import init_rerun

examples/phone_to_so100/record.py

@@ -15,11 +15,11 @@
 # limitations under the License.
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.datasets.feature_utils import combine_feature_dicts
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.datasets.pipeline_features import aggregate_pipeline_dataset_features, create_initial_features
-from lerobot.datasets.utils import combine_feature_dicts
 from lerobot.model.kinematics import RobotKinematics
-from lerobot.processor import RobotAction, RobotObservation, RobotProcessorPipeline
+from lerobot.processor import RobotProcessorPipeline
 from lerobot.processor.converters import (
     observation_to_transition,
     robot_action_observation_to_transition,
@@ -38,6 +38,7 @@ from lerobot.scripts.lerobot_record import record_loop
 from lerobot.teleoperators.phone.config_phone import PhoneConfig, PhoneOS
 from lerobot.teleoperators.phone.phone_processor import MapPhoneActionToRobotAction
 from lerobot.teleoperators.phone.teleop_phone import Phone
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.control_utils import init_keyboard_listener
 from lerobot.utils.utils import log_say
 from lerobot.utils.visualization_utils import init_rerun

examples/phone_to_so100/replay.py

@@ -18,7 +18,7 @@ import time
 
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.model.kinematics import RobotKinematics
-from lerobot.processor import RobotAction, RobotObservation, RobotProcessorPipeline
+from lerobot.processor import RobotProcessorPipeline
 from lerobot.processor.converters import (
     robot_action_observation_to_transition,
     transition_to_robot_action,
@@ -27,6 +27,7 @@ from lerobot.robots.so_follower import SO100Follower, SO100FollowerConfig
 from lerobot.robots.so_follower.robot_kinematic_processor import (
     InverseKinematicsEEToJoints,
 )
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.constants import ACTION
 from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
@@ -66,9 +67,7 @@ def main():
 
     # Fetch the dataset to replay
     dataset = LeRobotDataset(HF_REPO_ID, episodes=[EPISODE_IDX])
-    # Filter dataset to only include frames from the specified episode since episodes are chunked in dataset V3.0
-    episode_frames = dataset.hf_dataset.filter(lambda x: x["episode_index"] == EPISODE_IDX)
-    actions = episode_frames.select_columns(ACTION)
+    actions = dataset.select_columns(ACTION)
 
     # Connect to the robot
     robot.connect()
@@ -79,7 +78,7 @@ def main():
 
         print("Starting replay loop...")
         log_say(f"Replaying episode {EPISODE_IDX}")
-        for idx in range(len(episode_frames)):
+        for idx in range(dataset.num_frames):
             t0 = time.perf_counter()
 
             # Get recorded action from dataset

examples/phone_to_so100/teleoperate.py

@@ -16,7 +16,7 @@
 import time
 
 from lerobot.model.kinematics import RobotKinematics
-from lerobot.processor import RobotAction, RobotObservation, RobotProcessorPipeline
+from lerobot.processor import RobotProcessorPipeline
 from lerobot.processor.converters import (
     robot_action_observation_to_transition,
     transition_to_robot_action,
@@ -31,6 +31,7 @@ from lerobot.robots.so_follower.robot_kinematic_processor import (
 from lerobot.teleoperators.phone.config_phone import PhoneConfig, PhoneOS
 from lerobot.teleoperators.phone.phone_processor import MapPhoneActionToRobotAction
 from lerobot.teleoperators.phone.teleop_phone import Phone
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.visualization_utils import init_rerun, log_rerun_data
 
@@ -89,12 +90,13 @@ def main():
     teleop_device.connect()
 
     # Init rerun viewer
-    init_rerun(session_name="phone_so100_teleop")
+    init_rerun(session_name="phone_so100_teleop", robot=robot, reset_time=True)
 
     if not robot.is_connected or not teleop_device.is_connected:
         raise ValueError("Robot or teleop is not connected!")
 
     print("Starting teleop loop. Move your phone to teleoperate the robot...")
+    start = time.perf_counter()
     while True:
         t0 = time.perf_counter()
 
@@ -111,7 +113,7 @@ def main():
         _ = robot.send_action(joint_action)
 
         # Visualize
-        log_rerun_data(observation=phone_obs, action=joint_action)
+        log_rerun_data(observation=phone_obs, action=joint_action, log_time=time.perf_counter() - start)
 
         precise_sleep(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
 

examples/port_datasets/port_droid.py

@@ -22,7 +22,8 @@ from pathlib import Path
 import numpy as np
 import tensorflow_datasets as tfds
 
-from lerobot.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.utils.utils import get_elapsed_time_in_days_hours_minutes_seconds
 
 DROID_SHARDS = 2048

examples/port_datasets/slurm_upload.py

@@ -26,7 +26,7 @@ from huggingface_hub import HfApi
 from huggingface_hub.constants import REPOCARD_NAME
 from port_droid import DROID_SHARDS
 
-from lerobot.datasets.lerobot_dataset import CODEBASE_VERSION, LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import CODEBASE_VERSION, LeRobotDatasetMetadata
 from lerobot.datasets.utils import create_lerobot_dataset_card
 from lerobot.utils.utils import init_logging
 
@@ -155,7 +155,7 @@ class UploadDataset(PipelineStep):
         from datasets.utils.tqdm import disable_progress_bars
         from huggingface_hub import CommitOperationAdd, preupload_lfs_files
 
-        from lerobot.datasets.lerobot_dataset import LeRobotDatasetMetadata
+        from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
         from lerobot.utils.utils import init_logging
 
         init_logging()

examples/rtc/eval_dataset.py

@@ -27,8 +27,8 @@ measuring consistency and ground truth alignment.
 Usage:
     # Basic usage with smolvla policy
     uv run python examples/rtc/eval_dataset.py \
-        --policy.path=helper2424/smolvla_check_rtc_last3 \
-        --dataset.repo_id=helper2424/check_rtc \
+        --policy.path=<USER>/smolvla_check_rtc_last3 \
+        --dataset.repo_id=<USER>/check_rtc \
         --rtc.execution_horizon=8 \
         --device=mps \
         --rtc.max_guidance_weight=10.0 \
@@ -58,16 +58,16 @@ Usage:
         --device=cuda
 
     uv run python examples/rtc/eval_dataset.py \
-        --policy.path=lipsop/reuben_pi0 \
-        --dataset.repo_id=ReubenLim/so101_cube_in_cup \
+        --policy.path=<USER>/reuben_pi0 \
+        --dataset.repo_id=<USER>/so101_cube_in_cup \
         --rtc.execution_horizon=8 \
         --device=cuda
 
     # With torch.compile for faster inference (PyTorch 2.0+)
     # Note: CUDA graphs disabled by default due to in-place ops in denoising loop
     uv run python examples/rtc/eval_dataset.py \
-        --policy.path=helper2424/smolvla_check_rtc_last3 \
-        --dataset.repo_id=helper2424/check_rtc \
+        --policy.path=<USER>/smolvla_check_rtc_last3 \
+        --dataset.repo_id=<USER>/check_rtc \
         --rtc.execution_horizon=8 \
         --device=mps \
         --use_torch_compile=true \
@@ -75,8 +75,8 @@ Usage:
 
     # With torch.compile on CUDA (CUDA graphs disabled by default)
     uv run python examples/rtc/eval_dataset.py \
-        --policy.path=helper2424/smolvla_check_rtc_last3 \
-        --dataset.repo_id=helper2424/check_rtc \
+        --policy.path=<USER>/smolvla_check_rtc_last3 \
+        --dataset.repo_id=<USER>/check_rtc \
         --rtc.execution_horizon=8 \
         --device=cuda \
         --use_torch_compile=true \
@@ -84,8 +84,8 @@ Usage:
 
     # Enable CUDA graphs (advanced - may cause tensor aliasing errors)
     uv run python examples/rtc/eval_dataset.py \
-        --policy.path=helper2424/smolvla_check_rtc_last3 \
-        --dataset.repo_id=helper2424/check_rtc \
+        --policy.path=<USER>/smolvla_check_rtc_last3 \
+        --dataset.repo_id=<USER>/check_rtc \
         --use_torch_compile=true \
         --torch_compile_backend=inductor \
         --torch_compile_mode=max-autotune \
@@ -113,8 +113,9 @@ from lerobot.configs import parser
 from lerobot.configs.default import DatasetConfig
 from lerobot.configs.policies import PreTrainedConfig
 from lerobot.configs.types import RTCAttentionSchedule
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
 from lerobot.datasets.factory import resolve_delta_timestamps
-from lerobot.datasets.lerobot_dataset import LeRobotDataset, LeRobotDatasetMetadata
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.policies.factory import get_policy_class, make_pre_post_processors
 from lerobot.policies.rtc.configuration_rtc import RTCConfig
 from lerobot.policies.rtc.debug_visualizer import RTCDebugVisualizer

examples/rtc/eval_with_real_robot.py

@@ -28,7 +28,7 @@ For simulation environments, see eval_with_simulation.py
 Usage:
     # Run RTC with Real robot with RTC
     uv run examples/rtc/eval_with_real_robot.py \
-        --policy.path=helper2424/smolvla_check_rtc_last3 \
+        --policy.path=<USER>/smolvla_check_rtc_last3 \
         --policy.device=mps \
         --rtc.enabled=true \
         --rtc.execution_horizon=20 \
@@ -41,7 +41,7 @@ Usage:
 
     # Run RTC with Real robot without RTC
     uv run examples/rtc/eval_with_real_robot.py \
-        --policy.path=helper2424/smolvla_check_rtc_last3 \
+        --policy.path=<USER>/smolvla_check_rtc_last3 \
         --policy.device=mps \
         --rtc.enabled=false \
         --robot.type=so100_follower \
@@ -53,7 +53,7 @@ Usage:
 
     # Run RTC with Real robot with pi0.5 policy
     uv run examples/rtc/eval_with_real_robot.py \
-        --policy.path=helper2424/pi05_check_rtc \
+        --policy.path=<USER>/pi05_check_rtc \
         --policy.device=mps \
         --rtc.enabled=true \
         --rtc.execution_horizon=20 \
@@ -63,6 +63,26 @@ Usage:
         --robot.cameras="{ gripper: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30}, front: {type: opencv, index_or_path: 1, width: 640, height: 480, fps: 30}}" \
         --task="Move green small object into the purple platform" \
         --duration=120
+
+    # Run RTC with bi_openarm_follower (dual-arm OpenArms) and pi0.5 policy
+    python examples/rtc/eval_with_real_robot.py \
+        --policy.path=lerobot-data-collection/folding_final \
+        --robot.type=bi_openarm_follower \
+        --robot.cameras='{left_wrist: {type: opencv, index_or_path: "/dev/video4", width: 1280, height: 720, fps: 30}, base: {type: opencv, index_or_path: "/dev/video2", width: 640, height: 480, fps: 30}, right_wrist: {type: opencv, index_or_path: "/dev/video0", width: 1280, height: 720, fps: 30}}' \
+        --robot.left_arm_config.port=can1 \
+        --robot.left_arm_config.side=left \
+        --robot.left_arm_config.can_interface=socketcan \
+        --robot.right_arm_config.port=can0 \
+        --robot.right_arm_config.side=right \
+        --robot.right_arm_config.can_interface=socketcan \
+        --task="Fold the T-shirt properly" \
+        --fps=30 \
+        --duration=2000 \
+        --rtc.enabled=true \
+        --rtc.execution_horizon=20 \
+        --rtc.max_guidance_weight=5.0 \
+        --rtc.prefix_attention_schedule=LINEAR \
+        --device=cuda
 """
 
 import logging
@@ -78,28 +98,38 @@ from torch import Tensor
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig  # noqa: F401
 from lerobot.cameras.realsense.configuration_realsense import RealSenseCameraConfig  # noqa: F401
+from lerobot.cameras.zmq.configuration_zmq import ZMQCameraConfig  # noqa: F401
 from lerobot.configs import parser
 from lerobot.configs.policies import PreTrainedConfig
 from lerobot.configs.types import RTCAttentionSchedule
-from lerobot.datasets.utils import build_dataset_frame, hw_to_dataset_features
+from lerobot.datasets.feature_utils import build_dataset_frame, hw_to_dataset_features
 from lerobot.policies.factory import get_policy_class, make_pre_post_processors
 from lerobot.policies.rtc.action_queue import ActionQueue
 from lerobot.policies.rtc.configuration_rtc import RTCConfig
 from lerobot.policies.rtc.latency_tracker import LatencyTracker
+from lerobot.processor import (
+    NormalizerProcessorStep,
+    RelativeActionsProcessorStep,
+    TransitionKey,
+    create_transition,
+)
 from lerobot.processor.factory import (
     make_default_robot_action_processor,
     make_default_robot_observation_processor,
 )
+from lerobot.processor.relative_action_processor import to_relative_actions
 from lerobot.rl.process import ProcessSignalHandler
 from lerobot.robots import (  # noqa: F401
     Robot,
     RobotConfig,
+    bi_openarm_follower,
     bi_so_follower,
     koch_follower,
     so_follower,
+    unitree_g1,
 )
 from lerobot.robots.utils import make_robot_from_config
-from lerobot.utils.constants import OBS_IMAGES
+from lerobot.utils.constants import OBS_IMAGES, OBS_STATE
 from lerobot.utils.hub import HubMixin
 from lerobot.utils.utils import init_logging
 
@@ -210,6 +240,35 @@ def is_image_key(k: str) -> bool:
     return k.startswith(OBS_IMAGES)
 
 
+def _reanchor_relative_rtc_prefix(
+    prev_actions_absolute: Tensor,
+    current_state: Tensor,
+    relative_step: RelativeActionsProcessorStep,
+    normalizer_step: NormalizerProcessorStep | None,
+    policy_device: torch.device | str,
+) -> Tensor:
+    """Convert absolute leftovers into model-space for relative-action RTC policies.
+
+    When a policy uses relative actions, the RTC prefix (leftover actions from
+    the previous chunk) is stored in absolute space. Before feeding it back to
+    the policy we need to re-express it relative to the *current* robot state
+    and then re-normalize.
+    """
+    state = current_state.detach().cpu()
+    if state.dim() == 1:
+        state = state.unsqueeze(0)
+
+    action_cpu = prev_actions_absolute.detach().cpu()
+    mask = relative_step._build_mask(action_cpu.shape[-1])
+    relative_actions = to_relative_actions(action_cpu, state, mask)
+
+    transition = create_transition(action=relative_actions)
+    if normalizer_step is not None:
+        transition = normalizer_step(transition)
+
+    return transition[TransitionKey.ACTION].to(policy_device)
+
+
 def get_actions(
     policy,
     robot: RobotWrapper,
@@ -235,7 +294,15 @@ def get_actions(
         fps = cfg.fps
         time_per_chunk = 1.0 / fps
 
-        dataset_features = hw_to_dataset_features(robot.observation_features(), "observation")
+        # Only keep .pos joints + camera streams if the policy was trained on positions,
+        # not the full pos/vel/torque state the robot exposes.
+        observation_features_hw = {
+            key: value
+            for key, value in robot.observation_features().items()
+            if key.endswith(".pos") or isinstance(value, tuple)
+        }
+
+        dataset_features = hw_to_dataset_features(observation_features_hw, "observation")
         policy_device = policy.config.device
 
         # Load preprocessor and postprocessor from pretrained files
@@ -253,6 +320,25 @@ def get_actions(
 
         logger.info("[GET_ACTIONS] Preprocessor/postprocessor loaded successfully with embedded stats")
 
+        relative_step = next(
+            (s for s in preprocessor.steps if isinstance(s, RelativeActionsProcessorStep) and s.enabled),
+            None,
+        )
+        normalizer_step = next(
+            (s for s in preprocessor.steps if isinstance(s, NormalizerProcessorStep)),
+            None,
+        )
+        if relative_step is not None:
+            if relative_step.action_names is None:
+                cfg_names = getattr(cfg.policy, "action_feature_names", None)
+                if cfg_names:
+                    relative_step.action_names = list(cfg_names)
+                else:
+                    relative_step.action_names = [
+                        k for k in robot.robot.action_features if k.endswith(".pos")
+                    ]
+            logger.info("[GET_ACTIONS] Relative actions enabled: will re-anchor RTC prefix")
+
         get_actions_threshold = cfg.action_queue_size_to_get_new_actions
 
         if not cfg.rtc.enabled:
@@ -295,6 +381,28 @@ def get_actions(
 
                 preproceseded_obs = preprocessor(obs_with_policy_features)
 
+                # Re-anchor leftover actions for relative-action policies.
+                # We need the *postprocessed* (absolute) leftover, not the original
+                # (normalized/relative) one that get_left_over() returns.
+                if (
+                    prev_actions is not None
+                    and relative_step is not None
+                    and OBS_STATE in obs_with_policy_features
+                ):
+                    with action_queue.lock:
+                        if action_queue.queue is not None:
+                            prev_actions_abs = action_queue.queue[action_queue.last_index :].clone()
+                        else:
+                            prev_actions_abs = None
+                    if prev_actions_abs is not None and prev_actions_abs.numel() > 0:
+                        prev_actions = _reanchor_relative_rtc_prefix(
+                            prev_actions_absolute=prev_actions_abs,
+                            current_state=obs_with_policy_features[OBS_STATE],
+                            relative_step=relative_step,
+                            normalizer_step=normalizer_step,
+                            policy_device=policy_device,
+                        )
+
                 # Generate actions WITH RTC
                 actions = policy.predict_action_chunk(
                     preproceseded_obs,
@@ -350,6 +458,8 @@ def actor_control(
     try:
         logger.info("[ACTOR] Starting actor thread")
 
+        action_keys = [k for k in robot.action_features() if k.endswith(".pos")]
+
         action_count = 0
         action_interval = 1.0 / cfg.fps
 
@@ -361,7 +471,7 @@ def actor_control(
 
             if action is not None:
                 action = action.cpu()
-                action_dict = {key: action[i].item() for i, key in enumerate(robot.action_features())}
+                action_dict = {key: action[i].item() for i, key in enumerate(action_keys)}
                 action_processed = robot_action_processor((action_dict, None))
                 robot.send_action(action_processed)
 

examples/so100_to_so100_EE/evaluate.py

@@ -16,15 +16,13 @@
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
 from lerobot.configs.types import FeatureType, PolicyFeature
+from lerobot.datasets.feature_utils import combine_feature_dicts
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.datasets.pipeline_features import aggregate_pipeline_dataset_features, create_initial_features
-from lerobot.datasets.utils import combine_feature_dicts
 from lerobot.model.kinematics import RobotKinematics
 from lerobot.policies.act.modeling_act import ACTPolicy
 from lerobot.policies.factory import make_pre_post_processors
 from lerobot.processor import (
-    RobotAction,
-    RobotObservation,
     RobotProcessorPipeline,
     make_default_teleop_action_processor,
 )
@@ -40,6 +38,7 @@ from lerobot.robots.so_follower.robot_kinematic_processor import (
     InverseKinematicsEEToJoints,
 )
 from lerobot.scripts.lerobot_record import record_loop
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.control_utils import init_keyboard_listener
 from lerobot.utils.utils import log_say
 from lerobot.utils.visualization_utils import init_rerun

examples/so100_to_so100_EE/record.py

@@ -16,11 +16,11 @@
 
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
+from lerobot.datasets.feature_utils import combine_feature_dicts
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.datasets.pipeline_features import aggregate_pipeline_dataset_features, create_initial_features
-from lerobot.datasets.utils import combine_feature_dicts
 from lerobot.model.kinematics import RobotKinematics
-from lerobot.processor import RobotAction, RobotObservation, RobotProcessorPipeline
+from lerobot.processor import RobotProcessorPipeline
 from lerobot.processor.converters import (
     observation_to_transition,
     robot_action_observation_to_transition,
@@ -35,6 +35,7 @@ from lerobot.robots.so_follower.robot_kinematic_processor import (
 )
 from lerobot.scripts.lerobot_record import record_loop
 from lerobot.teleoperators.so_leader import SO100Leader, SO100LeaderConfig
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.control_utils import init_keyboard_listener
 from lerobot.utils.utils import log_say
 from lerobot.utils.visualization_utils import init_rerun

examples/so100_to_so100_EE/replay.py

@@ -19,7 +19,7 @@ import time
 
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.model.kinematics import RobotKinematics
-from lerobot.processor import RobotAction, RobotObservation, RobotProcessorPipeline
+from lerobot.processor import RobotProcessorPipeline
 from lerobot.processor.converters import (
     robot_action_observation_to_transition,
     transition_to_robot_action,
@@ -28,6 +28,7 @@ from lerobot.robots.so_follower import SO100Follower, SO100FollowerConfig
 from lerobot.robots.so_follower.robot_kinematic_processor import (
     InverseKinematicsEEToJoints,
 )
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.constants import ACTION
 from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.utils import log_say
@@ -67,9 +68,7 @@ def main():
 
     # Fetch the dataset to replay
     dataset = LeRobotDataset(HF_REPO_ID, episodes=[EPISODE_IDX])
-    # Filter dataset to only include frames from the specified episode since episodes are chunked in dataset V3.0
-    episode_frames = dataset.hf_dataset.filter(lambda x: x["episode_index"] == EPISODE_IDX)
-    actions = episode_frames.select_columns(ACTION)
+    actions = dataset.select_columns(ACTION)
 
     # Connect to the robot
     robot.connect()
@@ -80,7 +79,7 @@ def main():
 
         print("Starting replay loop...")
         log_say(f"Replaying episode {EPISODE_IDX}")
-        for idx in range(len(episode_frames)):
+        for idx in range(dataset.num_frames):
             t0 = time.perf_counter()
 
             # Get recorded action from dataset

examples/so100_to_so100_EE/teleoperate.py

@@ -17,7 +17,7 @@
 import time
 
 from lerobot.model.kinematics import RobotKinematics
-from lerobot.processor import RobotAction, RobotObservation, RobotProcessorPipeline
+from lerobot.processor import RobotProcessorPipeline
 from lerobot.processor.converters import (
     robot_action_observation_to_transition,
     robot_action_to_transition,
@@ -30,6 +30,7 @@ from lerobot.robots.so_follower.robot_kinematic_processor import (
     InverseKinematicsEEToJoints,
 )
 from lerobot.teleoperators.so_leader import SO100Leader, SO100LeaderConfig
+from lerobot.types import RobotAction, RobotObservation
 from lerobot.utils.robot_utils import precise_sleep
 from lerobot.utils.visualization_utils import init_rerun, log_rerun_data
 
@@ -94,9 +95,10 @@ def main():
     leader.connect()
 
     # Init rerun viewer
-    init_rerun(session_name="so100_so100_EE_teleop")
+    init_rerun(session_name="so100_so100_EE_teleop", robot=follower, reset_time=True)
 
     print("Starting teleop loop...")
+    start = time.perf_counter()
     while True:
         t0 = time.perf_counter()
 
@@ -116,7 +118,9 @@ def main():
         _ = follower.send_action(follower_joints_act)
 
         # Visualize
-        log_rerun_data(observation=leader_ee_act, action=follower_joints_act)
+        log_rerun_data(
+            observation=leader_ee_act, action=follower_joints_act, log_time=time.perf_counter() - start
+        )
 
         precise_sleep(max(1.0 / FPS - (time.perf_counter() - t0), 0.0))
 

examples/training/train_policy.py

@@ -19,8 +19,9 @@ from pathlib import Path
 import torch
 
 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.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.feature_utils import dataset_to_policy_features
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.policies.diffusion.configuration_diffusion import DiffusionConfig
 from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
 from lerobot.policies.factory import make_pre_post_processors

examples/training/train_with_streaming.py

@@ -20,9 +20,9 @@ from pathlib import Path
 import torch
 
 from lerobot.configs.types import FeatureType
-from lerobot.datasets.lerobot_dataset import LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.feature_utils import dataset_to_policy_features
 from lerobot.datasets.streaming_dataset import StreamingLeRobotDataset
-from lerobot.datasets.utils import dataset_to_policy_features
 from lerobot.policies.act.configuration_act import ACTConfig
 from lerobot.policies.act.modeling_act import ACTPolicy
 from lerobot.policies.factory import make_pre_post_processors

examples/tutorial/act/act_training_example.py

@@ -5,8 +5,9 @@ from pathlib import Path
 import torch
 
 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.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.feature_utils import dataset_to_policy_features
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.policies.act.configuration_act import ACTConfig
 from lerobot.policies.act.modeling_act import ACTPolicy
 from lerobot.policies.factory import make_pre_post_processors

examples/tutorial/act/act_using_example.py

@@ -1,7 +1,7 @@
 import torch
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
-from lerobot.datasets.lerobot_dataset import LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
 from lerobot.policies.act.modeling_act import ACTPolicy
 from lerobot.policies.factory import make_pre_post_processors
 from lerobot.policies.utils import build_inference_frame, make_robot_action

examples/tutorial/diffusion/diffusion_training_example.py

@@ -5,8 +5,9 @@ from pathlib import Path
 import torch
 
 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.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.feature_utils import dataset_to_policy_features
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
 from lerobot.policies.diffusion.configuration_diffusion import DiffusionConfig
 from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
 from lerobot.policies.factory import make_pre_post_processors

examples/tutorial/diffusion/diffusion_using_example.py

@@ -1,7 +1,7 @@
 import torch
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
-from lerobot.datasets.lerobot_dataset import LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
 from lerobot.policies.diffusion.modeling_diffusion import DiffusionPolicy
 from lerobot.policies.factory import make_pre_post_processors
 from lerobot.policies.utils import build_inference_frame, make_robot_action

examples/tutorial/pi0/using_pi0_example.py

@@ -1,7 +1,7 @@
 import torch
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
-from lerobot.datasets.utils import hw_to_dataset_features
+from lerobot.datasets.feature_utils import hw_to_dataset_features
 from lerobot.policies.factory import make_pre_post_processors
 from lerobot.policies.pi0.modeling_pi0 import PI0Policy
 from lerobot.policies.utils import build_inference_frame, make_robot_action

examples/tutorial/rl/hilserl_example.py

@@ -6,8 +6,8 @@ from queue import Empty, Full
 import torch
 import torch.optim as optim
 
+from lerobot.datasets.feature_utils import hw_to_dataset_features
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
-from lerobot.datasets.utils import hw_to_dataset_features
 from lerobot.envs.configs import HILSerlProcessorConfig, HILSerlRobotEnvConfig
 from lerobot.policies.sac.configuration_sac import SACConfig
 from lerobot.policies.sac.modeling_sac import SACPolicy

examples/tutorial/smolvla/using_smolvla_example.py

@@ -1,7 +1,7 @@
 import torch
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig
-from lerobot.datasets.utils import hw_to_dataset_features
+from lerobot.datasets.feature_utils import hw_to_dataset_features
 from lerobot.policies.factory import make_pre_post_processors
 from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
 from lerobot.policies.utils import build_inference_frame, make_robot_action

pyproject.toml

@@ -25,11 +25,11 @@ discord = "https://discord.gg/s3KuuzsPFb"
 
 [project]
 name = "lerobot"
-version = "0.4.4"
+version = "0.5.1"
 description = "🤗 LeRobot: State-of-the-art Machine Learning for Real-World Robotics in Pytorch"
 dynamic = ["readme"]
 license = { text = "Apache-2.0" }
-requires-python = ">=3.10"
+requires-python = ">=3.12"
 authors = [
     { name = "Rémi Cadène", email = "re.cadene@gmail.com" },
     { name = "Simon Alibert", email = "alibert.sim@gmail.com" },
@@ -50,7 +50,8 @@ classifiers = [
     "Intended Audience :: Education",
     "Intended Audience :: Science/Research",
     "License :: OSI Approved :: Apache Software License",
-    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
     "Topic :: Software Development :: Build Tools",
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
 ]
@@ -59,28 +60,30 @@ keywords = ["lerobot", "huggingface", "robotics",  "machine learning", "artifici
 dependencies = [
 
     # Hugging Face dependencies
-    "datasets>=4.0.0,<4.2.0",
+    "datasets>=4.0.0,<5.0.0",
     "diffusers>=0.27.2,<0.36.0",
-    "huggingface-hub[hf-transfer,cli]>=0.34.2,<0.36.0",
+    "huggingface-hub>=1.0.0,<2.0.0",
     "accelerate>=1.10.0,<2.0.0",
 
     # Core dependencies
+    "numpy>=2.0.0,<2.3.0", # NOTE: Explicitly listing numpy helps the resolver converge faster. Upper bound imposed by opencv-python-headless.
     "setuptools>=71.0.0,<81.0.0",
     "cmake>=3.29.0.1,<4.2.0",
+    "packaging>=24.2,<26.0",
+
+    "torch>=2.2.1,<2.11.0",
+    "torchcodec>=0.2.1,<0.11.0; sys_platform != 'win32' and (sys_platform != 'linux' or (platform_machine != 'aarch64' and platform_machine != 'arm64' and platform_machine != 'armv7l')) and (sys_platform != 'darwin' or platform_machine != 'x86_64')",
+    "torchvision>=0.21.0,<0.26.0",
+
     "einops>=0.8.0,<0.9.0",
-    "opencv-python-headless>=4.9.0,<4.13.0",
+    "opencv-python-headless>=4.9.0,<4.14.0",
     "av>=15.0.0,<16.0.0",
     "jsonlines>=4.0.0,<5.0.0",
-    "packaging>=24.2,<26.0",
-    "pynput>=1.7.7,<1.9.0",
+    "pynput>=1.7.8,<1.9.0",
     "pyserial>=3.5,<4.0",
+
     "wandb>=0.24.0,<0.25.0",
-
-    "torch>=2.2.1,<2.8.0", # TODO: Bumb dependency
-    "torchcodec>=0.2.1,<0.6.0; sys_platform != 'win32' and (sys_platform != 'linux' or (platform_machine != 'aarch64' and platform_machine != 'arm64' and platform_machine != 'armv7l')) and (sys_platform != 'darwin' or platform_machine != 'x86_64')", # TODO: Bumb dependency
-    "torchvision>=0.21.0,<0.23.0", # TODO: Bumb dependency
-
-    "draccus==0.10.0", # TODO: Remove ==
+    "draccus==0.10.0", # TODO: Relax version constraint
     "gymnasium>=1.1.1,<2.0.0",
     "rerun-sdk>=0.24.0,<0.27.0",
 
@@ -95,14 +98,20 @@ dependencies = [
 
 # Common
 pygame-dep = ["pygame>=2.5.1,<2.7.0"]
-placo-dep = ["placo>=0.9.6,<0.10.0"]
-transformers-dep = ["transformers>=4.57.1,<5.0.0"]
+placo-dep = ["placo>=0.9.6,<0.9.17"]
+transformers-dep = ["transformers==5.3.0"] # TODO(Steven): https://github.com/huggingface/lerobot/pull/3249
 grpcio-dep = ["grpcio==1.73.1", "protobuf>=6.31.1,<6.32.0"]
+can-dep = ["python-can>=4.2.0,<5.0.0"]
+peft-dep = ["peft>=0.18.0,<1.0.0"]
+scipy-dep = ["scipy>=1.14.0,<2.0.0"]
+qwen-vl-utils-dep = ["qwen-vl-utils>=0.0.11,<0.1.0"]
+matplotlib-dep = ["matplotlib>=3.10.3,<4.0.0", "contourpy>=1.3.0,<2.0.0"] # NOTE: Explicitly listing contourpy helps the resolver converge faster.
 
 # Motors
 feetech = ["feetech-servo-sdk>=1.0.0,<2.0.0"]
 dynamixel = ["dynamixel-sdk>=3.7.31,<3.9.0"]
-damiao = ["python-can>=4.2.0,<5.0.0"]
+damiao = ["lerobot[can-dep]"]
+robstride = ["lerobot[can-dep]"]
 
 # Robots
 openarms = ["lerobot[damiao]"]
@@ -110,34 +119,36 @@ gamepad = ["lerobot[pygame-dep]", "hidapi>=0.14.0,<0.15.0"]
 hopejr = ["lerobot[feetech]", "lerobot[pygame-dep]"]
 lekiwi = ["lerobot[feetech]", "pyzmq>=26.2.1,<28.0.0"]
 unitree_g1 = [
+    # "unitree-sdk2==1.0.1",
     "pyzmq>=26.2.1,<28.0.0",
     "onnxruntime>=1.16.0,<2.0.0",
-    "pin>=3.0.0,<4.0.0",
+    "onnx>=1.16.0,<2.0.0",
     "meshcat>=0.3.0,<0.4.0",
-    "matplotlib>=3.9.0,<4.0.0",
-    "casadi>=3.6.0,<4.0.0",
+    "lerobot[matplotlib-dep]",
+    "lerobot[pygame-dep]",
 ]
 reachy2 = ["reachy2_sdk>=1.0.15,<1.1.0"]
 kinematics = ["lerobot[placo-dep]"]
 intelrealsense = [
     "pyrealsense2>=2.55.1.6486,<2.57.0 ; sys_platform != 'darwin'",
-    "pyrealsense2-macosx>=2.54,<2.55.0 ; sys_platform == 'darwin'",
+    "pyrealsense2-macosx>=2.54,<2.57.0 ; sys_platform == 'darwin'",
 ]
-phone = ["hebi-py>=2.8.0,<2.12.0", "teleop>=0.1.0,<0.2.0", "fastapi<1.0"]
+phone = ["hebi-py>=2.8.0,<2.12.0", "teleop>=0.1.0,<0.2.0", "fastapi<1.0", "lerobot[scipy-dep]"]
 
 # Policies
 wallx = [
-    "transformers==4.49.0",
-    "peft==0.17.1",
-    "scipy==1.15.3",
-    "torchdiffeq==0.2.5",
-    "qwen_vl_utils==0.0.11"
+    "lerobot[transformers-dep]",
+    "lerobot[peft]",
+    "lerobot[scipy-dep]",
+    "torchdiffeq>=0.2.4,<0.3.0",
+    "lerobot[qwen-vl-utils-dep]",
 ]
-pi = ["transformers @ git+https://github.com/huggingface/transformers.git@fix/lerobot_openpi", "scipy>=1.10.1,<1.15"]
+pi = ["lerobot[transformers-dep]", "lerobot[scipy-dep]"]
 smolvla = ["lerobot[transformers-dep]", "num2words>=0.5.14,<0.6.0", "accelerate>=1.7.0,<2.0.0", "safetensors>=0.4.3,<1.0.0"]
+multi_task_dit = ["lerobot[transformers-dep]"]
 groot = [
     "lerobot[transformers-dep]",
-    "peft>=0.13.0,<1.0.0",
+    "lerobot[peft]",
     "dm-tree>=0.1.8,<1.0.0",
     "timm>=1.0.0,<1.1.0",
     "safetensors>=0.4.3,<1.0.0",
@@ -146,13 +157,14 @@ groot = [
     "ninja>=1.11.1,<2.0.0",
     "flash-attn>=2.5.9,<3.0.0 ; sys_platform != 'darwin'"
 ]
-sarm = ["lerobot[transformers-dep]", "faker>=33.0.0,<35.0.0", "matplotlib>=3.10.3,<4.0.0", "qwen-vl-utils>=0.0.14,<0.1.0"]
+sarm = ["lerobot[transformers-dep]", "faker>=33.0.0,<35.0.0", "lerobot[matplotlib-dep]", "lerobot[qwen-vl-utils-dep]"]
 xvla = ["lerobot[transformers-dep]"]
 hilserl = ["lerobot[transformers-dep]", "gym-hil>=0.1.13,<0.2.0", "lerobot[grpcio-dep]", "lerobot[placo-dep]"]
 
 # Features
-async = ["lerobot[grpcio-dep]", "matplotlib>=3.10.3,<4.0.0"]
-peft = ["lerobot[transformers-dep]", "peft>=0.18.0,<1.0.0"]
+async = ["lerobot[grpcio-dep]", "lerobot[matplotlib-dep]"]
+peft = ["lerobot[transformers-dep]", "lerobot[peft-dep]"]
+audio = ["sounddevice>=0.5.1,<0.6.0", "soundfile>=0.13.1,<0.14.0", "librosa>=0.11.0,<0.12.0", "torchaudio>=2.6.0,<2.10.0"]
 
 # Development
 dev = ["pre-commit>=3.7.0,<5.0.0", "debugpy>=1.8.1,<1.9.0", "lerobot[grpcio-dep]", "grpcio-tools==1.73.1", "mypy>=1.19.1"]
@@ -160,13 +172,19 @@ test = ["pytest>=8.1.0,<9.0.0", "pytest-timeout>=2.4.0,<3.0.0", "pytest-cov>=5.0
 video_benchmark = ["scikit-image>=0.23.2,<0.26.0", "pandas>=2.2.2,<2.4.0"]
 
 # Simulation
-aloha = ["gym-aloha>=0.1.2,<0.2.0"]
+# NOTE: Explicitly listing scipy helps flatten the dependecy tree.
+aloha = ["gym-aloha>=0.1.2,<0.2.0", "lerobot[scipy-dep]"]
 pusht = ["gym-pusht>=0.1.5,<0.2.0", "pymunk>=6.6.0,<7.0.0"] # TODO: Fix pymunk version in gym-pusht instead
-libero = ["lerobot[transformers-dep]", "hf-libero>=0.1.3,<0.2.0"]
-metaworld = ["metaworld==3.0.0"]
+libero = ["lerobot[transformers-dep]", "hf-libero>=0.1.3,<0.2.0; sys_platform == 'linux'", "lerobot[scipy-dep]"]
+metaworld = ["metaworld==3.0.0", "lerobot[scipy-dep]"]
 
 # All
 all = [
+    # NOTE(resolver hint): scipy is pulled in transitively via lerobot[scipy-dep] through
+    # multiple extras (aloha, metaworld, pi, wallx, phone). Listing it explicitly
+    # helps pip's resolver converge by constraining scipy early, before it encounters
+    # the loose scipy requirements from transitive deps like dm-control and metaworld.
+    "scipy>=1.14.0,<2.0.0",
     "lerobot[dynamixel]",
     "lerobot[gamepad]",
     "lerobot[hopejr]",
@@ -174,23 +192,25 @@ all = [
     "lerobot[reachy2]",
     "lerobot[kinematics]",
     "lerobot[intelrealsense]",
-    # "lerobot[wallx]",
-    # "lerobot[pi]", TODO(Pepijn): Update pi to transformers v5
+    "lerobot[wallx]",
+    "lerobot[pi]",
     "lerobot[smolvla]",
     # "lerobot[groot]", TODO(Steven): Gr00t requires specific installation instructions for flash-attn
     "lerobot[xvla]",
     "lerobot[hilserl]",
     "lerobot[async]",
+    "lerobot[audio]",
     "lerobot[dev]",
     "lerobot[test]",
     "lerobot[video_benchmark]",
     "lerobot[aloha]",
     "lerobot[pusht]",
     "lerobot[phone]",
-    "lerobot[libero]",
+    "lerobot[libero]; sys_platform == 'linux'",
     "lerobot[metaworld]",
     "lerobot[sarm]",
     "lerobot[peft]",
+    # "lerobot[unitree_g1]", TODO: Unitree requires specific installation instructions for unitree_sdk2
 ]
 
 [project.scripts]
@@ -212,11 +232,14 @@ lerobot-edit-dataset="lerobot.scripts.lerobot_edit_dataset:main"
 lerobot-setup-can="lerobot.scripts.lerobot_setup_can:main"
 
 # ---------------- Tool Configurations ----------------
+[tool.setuptools.package-data]
+lerobot = ["envs/*.json"]
+
 [tool.setuptools.packages.find]
 where = ["src"]
 
 [tool.ruff]
-target-version = "py310"
+target-version = "py312"
 line-length = 110
 exclude = ["tests/artifacts/**/*.safetensors", "*_pb2.py", "*_pb2_grpc.py"]
 
@@ -308,7 +331,7 @@ default.extend-ignore-identifiers-re = [
 # Uncomment [tool.mypy] first, then uncomment individual module overrides as they get proper type annotations
 
 [tool.mypy]
-python_version = "3.10"
+python_version = "3.12"
 ignore_missing_imports = true
 follow_imports = "skip"
 # warn_return_any = true
@@ -392,85 +415,3 @@ ignore_errors = false
 # [[tool.mypy.overrides]]
 # module = "lerobot.scripts.*"
 # ignore_errors = false
-
-[tool.uv]
-# wallx requires transformers==4.49.0 which conflicts with other extras that need >=4.53.0
-conflicts = [
-    [
-        { extra = "wallx" },
-        { extra = "transformers-dep" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "pi" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "smolvla" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "groot" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "xvla" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "sarm" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "hilserl" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "libero" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "peft" },
-    ],
-    [
-        { extra = "wallx" },
-        { extra = "all" },
-    ],
-    # pi uses custom branch which conflicts with transformers-dep
-    [
-        { extra = "pi" },
-        { extra = "transformers-dep" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "smolvla" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "groot" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "xvla" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "sarm" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "hilserl" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "libero" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "peft" },
-    ],
-    [
-        { extra = "pi" },
-        { extra = "all" },
-    ],
-]

requirements-macos.txt

@@ -1,76 +1,73 @@
 #
-# This file is autogenerated by pip-compile with Python 3.10
+# This file is autogenerated by pip-compile with Python 3.12
 # by the following command:
 #
 #    pip-compile --output-file=requirements-macos.txt requirements.in
 #
 -e .[all]
     # via -[all]
-absl-py==2.3.1
+absl-py==2.4.0
     # via
     #   dm-control
     #   dm-env
     #   dm-tree
     #   labmaze
     #   mujoco
-    #   tensorboard
-accelerate==1.11.0
+accelerate==1.13.0
     # via
     #   lerobot
     #   peft
 aiohappyeyeballs==2.6.1
     # via aiohttp
-aiohttp==3.13.1
+aiohttp==3.13.3
     # via fsspec
 aiosignal==1.4.0
     # via aiohttp
+annotated-doc==0.0.4
+    # via
+    #   fastapi
+    #   typer
 annotated-types==0.7.0
     # via pydantic
-antlr4-python3-runtime==4.9.3
-    # via
-    #   hydra-core
-    #   omegaconf
-anyio==4.11.0
+anyio==4.12.1
     # via
+    #   httpx
     #   starlette
     #   watchfiles
-asttokens==3.0.0
+asttokens==3.0.1
     # via stack-data
-async-timeout==5.0.1
-    # via aiohttp
 attrs==25.4.0
     # via
     #   aiohttp
     #   dm-tree
     #   jsonlines
-    #   jsonschema
-    #   referencing
     #   rerun-sdk
 av==15.1.0
-    # via lerobot
-bddl==1.0.1
-    # via libero
-certifi==2025.10.5
     # via
+    #   lerobot
+    #   qwen-vl-utils
+certifi==2026.2.25
+    # via
+    #   httpcore
+    #   httpx
     #   requests
     #   sentry-sdk
 cffi==2.0.0
     # via pymunk
-cfgv==3.4.0
+cfgv==3.5.0
     # via pre-commit
-charset-normalizer==3.4.4
+charset-normalizer==3.4.5
     # via requests
-click==8.3.0
+click==8.3.1
     # via
+    #   typer
     #   uvicorn
     #   wandb
-cloudpickle==3.1.1
-    # via
-    #   gymnasium
-    #   libero
-cmake==4.1.0
+cloudpickle==3.1.2
+    # via gymnasium
+cmake==4.1.3
     # via lerobot
-cmeel==0.57.3
+cmeel==0.59.0
     # via
     #   cmeel-assimp
     #   cmeel-boost
@@ -108,15 +105,17 @@ cmeel-zlib==1.3.1
     # via cmeel-assimp
 coal-library==3.0.1
     # via pin
-contourpy==1.3.2
-    # via matplotlib
-coverage[toml]==7.11.0
+contourpy==1.3.3
+    # via
+    #   lerobot
+    #   matplotlib
+coverage[toml]==7.13.4
     # via pytest-cov
 cycler==0.12.1
     # via matplotlib
-datasets==4.1.1
+datasets==4.6.1
     # via lerobot
-debugpy==1.8.17
+debugpy==1.8.20
     # via lerobot
 decorator==5.2.1
     # via ipython
@@ -130,7 +129,7 @@ dill==0.4.0
     #   multiprocess
 distlib==0.4.0
     # via virtualenv
-dm-control==1.0.34
+dm-control==1.0.37
     # via gym-aloha
 dm-env==1.6
     # via dm-control
@@ -138,69 +137,55 @@ dm-tree==0.1.9
     # via
     #   dm-control
     #   dm-env
-    #   lerobot
 docopt==0.6.2
     # via num2words
 draccus==0.10.0
     # via lerobot
 dynamixel-sdk==3.8.4
     # via lerobot
-easydict==1.13
-    # via libero
-egl-probe @ git+https://github.com/huggingface/egl_probe.git
-    # via
-    #   libero
-    #   robomimic
 eigenpy==3.10.3
     # via coal-library
-einops==0.8.1
-    # via
-    #   lerobot
-    #   libero
+einops==0.8.2
+    # via lerobot
 eiquadprog==1.2.9
     # via placo
-etils[epath,epy]==1.13.0
+etils[epath,epy]==1.14.0
     # via mujoco
-exceptiongroup==1.3.0
-    # via
-    #   anyio
-    #   ipython
-    #   pytest
 executing==2.2.1
     # via stack-data
+faker==34.0.2
+    # via lerobot
 farama-notifications==0.0.4
     # via gymnasium
-fastapi==0.119.1
-    # via teleop
-fastjsonschema==2.21.2
-    # via nbformat
+fastapi==0.135.1
+    # via
+    #   lerobot
+    #   teleop
 feetech-servo-sdk==1.0.0
     # via lerobot
-filelock==3.20.0
+filelock==3.25.0
     # via
     #   datasets
     #   diffusers
     #   huggingface-hub
+    #   python-discovery
     #   torch
-    #   transformers
     #   virtualenv
-fonttools==4.60.1
+fonttools==4.61.1
     # via matplotlib
 frozenlist==1.8.0
     # via
     #   aiohttp
     #   aiosignal
-fsspec[http]==2025.9.0
+fsspec[http]==2026.2.0
     # via
     #   datasets
     #   etils
     #   huggingface-hub
     #   torch
-future==1.0.0
-    # via libero
 gitdb==4.0.12
     # via gitpython
-gitpython==3.1.45
+gitpython==3.1.46
     # via wandb
 glfw==2.10.0
     # via
@@ -212,7 +197,6 @@ grpcio==1.73.1
     #   lerobot
     #   reachy2-sdk
     #   reachy2-sdk-api
-    #   tensorboard
 grpcio-tools==1.73.1
     # via
     #   lerobot
@@ -223,71 +207,67 @@ gym-hil==0.1.13
     # via lerobot
 gym-pusht==0.1.6
     # via lerobot
-gymnasium==1.2.1
+gymnasium==1.2.3
     # via
     #   gym-aloha
     #   gym-hil
     #   gym-pusht
     #   lerobot
-    #   libero
     #   metaworld
 h11==0.16.0
-    # via uvicorn
-h5py==3.15.1
-    # via robomimic
+    # via
+    #   httpcore
+    #   uvicorn
 hebi-py==2.11.0
     # via lerobot
-hf-transfer==0.1.9
-    # via huggingface-hub
-hf-xet==1.1.10
+hf-xet==1.3.2
     # via huggingface-hub
 hidapi==0.14.0.post4
     # via
     #   gym-hil
     #   lerobot
+httpcore==1.0.9
+    # via httpx
 httptools==0.7.1
     # via uvicorn
-huggingface-hub[cli,hf-transfer]==0.35.3
+httpx==0.28.1
+    # via
+    #   datasets
+    #   huggingface-hub
+huggingface-hub==1.6.0
     # via
     #   accelerate
     #   datasets
     #   diffusers
     #   lerobot
     #   peft
-    #   timm
     #   tokenizers
     #   transformers
-hydra-core==1.3.2
-    # via libero
-identify==2.6.15
+identify==2.6.17
     # via pre-commit
 idna==3.11
     # via
     #   anyio
+    #   httpx
     #   requests
     #   yarl
-imageio[ffmpeg]==2.37.0
+imageio[ffmpeg]==2.37.2
     # via
     #   gym-aloha
     #   gym-hil
     #   lerobot
     #   metaworld
-    #   robomimic
     #   scikit-image
 imageio-ffmpeg==0.6.0
-    # via
-    #   imageio
-    #   robomimic
-importlib-metadata==8.7.0
+    # via imageio
+importlib-metadata==8.7.1
     # via diffusers
-importlib-resources==6.5.2
-    # via etils
 iniconfig==2.3.0
     # via pytest
-inquirerpy==0.3.4
-    # via huggingface-hub
-ipython==8.37.0
+ipython==9.11.0
     # via meshcat
+ipython-pygments-lexers==1.1.1
+    # via ipython
 ischedule==1.2.7
     # via placo
 jedi==0.19.2
@@ -296,44 +276,24 @@ jinja2==3.1.6
     # via torch
 jsonlines==4.0.0
     # via lerobot
-jsonschema==4.25.1
-    # via nbformat
-jsonschema-specifications==2025.9.1
-    # via jsonschema
-jupyter-core==5.9.1
-    # via nbformat
-jupytext==1.18.1
-    # via bddl
 kiwisolver==1.4.9
     # via matplotlib
 labmaze==1.0.6
     # via dm-control
-lazy-loader==0.4
+lazy-loader==0.5
     # via scikit-image
-libero @ git+https://github.com/huggingface/lerobot-libero.git@main
-    # via lerobot
-llvmlite==0.45.1
-    # via numba
+librt==0.8.1
+    # via mypy
 lxml==6.0.2
     # via dm-control
-markdown==3.9
-    # via tensorboard
 markdown-it-py==4.0.0
-    # via
-    #   jupytext
-    #   mdit-py-plugins
+    # via rich
 markupsafe==3.0.3
-    # via
-    #   jinja2
-    #   werkzeug
-matplotlib==3.10.7
-    # via
-    #   lerobot
-    #   libero
+    # via jinja2
+matplotlib==3.10.8
+    # via lerobot
 matplotlib-inline==0.2.1
     # via ipython
-mdit-py-plugins==0.5.0
-    # via jupytext
 mdurl==0.1.2
     # via markdown-it-py
 mergedeep==1.3.4
@@ -346,41 +306,35 @@ mock-serial==0.0.1
     # via lerobot
 mpmath==1.3.0
     # via sympy
-mujoco==3.3.7
+mujoco==3.5.0
     # via
     #   dm-control
     #   gym-aloha
     #   gym-hil
-    #   libero
     #   metaworld
-    #   robosuite
-multidict==6.7.0
+multidict==6.7.1
     # via
     #   aiohttp
     #   yarl
-multiprocess==0.70.16
+multiprocess==0.70.18
     # via datasets
+mypy==1.19.1
+    # via lerobot
 mypy-extensions==1.1.0
-    # via typing-inspect
-nbformat==5.10.4
-    # via jupytext
-networkx==3.4.2
     # via
-    #   bddl
+    #   mypy
+    #   typing-inspect
+networkx==3.6.1
+    # via
     #   scikit-image
     #   torch
-ninja==1.13.0
-    # via lerobot
-nodeenv==1.9.1
+nodeenv==1.10.0
     # via pre-commit
 num2words==0.5.14
     # via lerobot
-numba==0.62.1
-    # via robosuite
 numpy==2.2.6
     # via
     #   accelerate
-    #   bddl
     #   cmeel-boost
     #   contourpy
     #   datasets
@@ -389,16 +343,14 @@ numpy==2.2.6
     #   dm-env
     #   dm-tree
     #   gymnasium
-    #   h5py
     #   hebi-py
     #   imageio
     #   labmaze
-    #   libero
+    #   lerobot
     #   matplotlib
     #   meshcat
     #   metaworld
     #   mujoco
-    #   numba
     #   opencv-python
     #   opencv-python-headless
     #   pandas
@@ -406,26 +358,18 @@ numpy==2.2.6
     #   pyquaternion
     #   reachy2-sdk
     #   rerun-sdk
-    #   robomimic
-    #   robosuite
     #   scikit-image
     #   scipy
     #   shapely
     #   teleop
-    #   tensorboard
-    #   tensorboardx
     #   tifffile
     #   torchvision
     #   transformers
     #   transforms3d
-omegaconf==2.3.0
-    # via hydra-core
-opencv-python==4.12.0.88
+opencv-python==4.13.0.92
     # via
     #   gym-pusht
-    #   libero
     #   reachy2-sdk
-    #   robosuite
 opencv-python-headless==4.12.0.88
     # via lerobot
 orderly-set==5.5.0
@@ -435,97 +379,87 @@ packaging==25.0
     #   accelerate
     #   datasets
     #   huggingface-hub
-    #   hydra-core
-    #   jupytext
     #   lazy-loader
     #   lerobot
     #   matplotlib
     #   peft
     #   pytest
+    #   qwen-vl-utils
     #   reachy2-sdk
     #   scikit-image
-    #   tensorboard
-    #   tensorboardx
     #   transformers
     #   wandb
 pandas==2.3.3
     # via
     #   datasets
     #   lerobot
-parso==0.8.5
+parso==0.8.6
     # via jedi
-peft==0.17.1
+pathspec==1.0.4
+    # via mypy
+peft==0.18.1
     # via lerobot
 pexpect==4.9.0
     # via ipython
-pfzy==0.3.4
-    # via inquirerpy
-pillow==12.0.0
+pillow==12.1.1
     # via
     #   diffusers
     #   imageio
-    #   lerobot
     #   matplotlib
     #   meshcat
+    #   qwen-vl-utils
     #   rerun-sdk
-    #   robosuite
     #   scikit-image
-    #   tensorboard
     #   torchvision
 pin==3.4.0
     # via placo
-placo==0.9.14
+placo==0.9.16
     # via lerobot
-platformdirs==4.5.0
+platformdirs==4.9.4
     # via
-    #   jupyter-core
+    #   python-discovery
     #   virtualenv
     #   wandb
 pluggy==1.6.0
     # via
     #   pytest
     #   pytest-cov
-pre-commit==4.3.0
+pre-commit==4.5.1
     # via lerobot
 prompt-toolkit==3.0.52
-    # via
-    #   inquirerpy
-    #   ipython
+    # via ipython
 propcache==0.4.1
     # via
     #   aiohttp
     #   yarl
-protobuf==6.31.0
+protobuf==6.31.1
     # via
     #   dm-control
     #   grpcio-tools
     #   lerobot
     #   reachy2-sdk
     #   reachy2-sdk-api
-    #   tensorboard
-    #   tensorboardx
     #   wandb
-psutil==7.1.1
+psutil==7.2.2
     # via
     #   accelerate
     #   imageio
     #   peft
-    #   robomimic
 ptyprocess==0.7.0
     # via pexpect
 pure-eval==0.2.3
     # via stack-data
-pyarrow==21.0.0
+pyarrow==23.0.1
     # via
     #   datasets
     #   rerun-sdk
-pycparser==2.23
+pycparser==3.0
     # via cffi
-pydantic==2.12.3
+pydantic==2.12.5
     # via
     #   fastapi
     #   wandb
-pydantic-core==2.41.4
+pydantic-core==2.41.5
     # via pydantic
 pygame==2.6.1
     # via
@@ -535,33 +469,35 @@ pygame==2.6.1
 pygments==2.19.2
     # via
     #   ipython
+    #   ipython-pygments-lexers
     #   pytest
+    #   rich
 pymunk==6.11.1
     # via
     #   gym-pusht
     #   lerobot
-pyngrok==7.4.1
+pyngrok==7.5.1
     # via meshcat
 pynput==1.8.1
     # via
     #   gym-hil
     #   lerobot
-pyobjc-core==12.0
+pyobjc-core==12.1
     # via
     #   pyobjc-framework-applicationservices
     #   pyobjc-framework-cocoa
     #   pyobjc-framework-coretext
     #   pyobjc-framework-quartz
-pyobjc-framework-applicationservices==12.0
+pyobjc-framework-applicationservices==12.1
     # via pynput
-pyobjc-framework-cocoa==12.0
+pyobjc-framework-cocoa==12.1
     # via
     #   pyobjc-framework-applicationservices
     #   pyobjc-framework-coretext
     #   pyobjc-framework-quartz
-pyobjc-framework-coretext==12.0
+pyobjc-framework-coretext==12.1
     # via pyobjc-framework-applicationservices
-pyobjc-framework-quartz==12.0
+pyobjc-framework-quartz==12.1
     # via
     #   pynput
     #   pyobjc-framework-applicationservices
@@ -570,13 +506,13 @@ pyopengl==3.1.10
     # via
     #   dm-control
     #   mujoco
-pyparsing==3.2.5
+pyparsing==3.3.2
     # via
     #   dm-control
     #   matplotlib
 pyquaternion==0.9.9
     # via reachy2-sdk
-pyrealsense2-macosx==2.54.2
+pyrealsense2-macosx==2.56.5
     # via lerobot
 pyserial==3.5
     # via
@@ -585,7 +521,6 @@ pyserial==3.5
     #   lerobot
 pytest==8.4.2
     # via
-    #   bddl
     #   lerobot
     #   pytest-cov
     #   pytest-timeout
@@ -596,11 +531,14 @@ pytest-timeout==2.4.0
     # via lerobot
 python-dateutil==2.9.0.post0
     # via
+    #   faker
     #   matplotlib
     #   pandas
-python-dotenv==1.1.1
+python-discovery==1.1.1
+    # via virtualenv
+python-dotenv==1.2.2
     # via uvicorn
-pytz==2025.2
+pytz==2026.1.post1
     # via pandas
 pyyaml==6.0.3
     # via
@@ -609,13 +547,10 @@ pyyaml==6.0.3
     #   draccus
     #   hebi-py
     #   huggingface-hub
-    #   jupytext
-    #   omegaconf
     #   peft
     #   pre-commit
     #   pyngrok
     #   pyyaml-include
-    #   timm
     #   transformers
     #   uvicorn
     #   wandb
@@ -625,15 +560,13 @@ pyzmq==27.1.0
     # via
     #   lerobot
     #   meshcat
-reachy2-sdk==1.0.14
+qwen-vl-utils==0.0.14
+    # via lerobot
+reachy2-sdk==1.0.15
     # via lerobot
 reachy2-sdk-api==1.0.21
     # via reachy2-sdk
-referencing==0.37.0
-    # via
-    #   jsonschema
-    #   jsonschema-specifications
-regex==2025.10.23
+regex==2026.2.28
     # via
     #   diffusers
     #   transformers
@@ -642,184 +575,150 @@ requests==2.32.5
     #   datasets
     #   diffusers
     #   dm-control
-    #   huggingface-hub
+    #   qwen-vl-utils
     #   teleop
-    #   transformers
     #   wandb
-rerun-sdk==0.26.1
+rerun-sdk==0.26.2
     # via lerobot
 rhoban-cmeel-jsoncpp==1.9.4.9
     # via placo
-robomimic==0.2.0
-    # via libero
-robosuite==1.4.0
-    # via libero
-rpds-py==0.28.0
-    # via
-    #   jsonschema
-    #   referencing
-safetensors==0.6.2
+rich==14.3.3
+    # via typer
+safetensors==0.7.0
     # via
     #   accelerate
     #   diffusers
     #   lerobot
     #   peft
-    #   timm
     #   transformers
 scikit-image==0.25.2
     # via
     #   gym-pusht
     #   lerobot
-scipy==1.15.3
+scipy==1.17.1
     # via
     #   dm-control
+    #   lerobot
     #   metaworld
-    #   robosuite
     #   scikit-image
-sentry-sdk==2.42.1
+    #   torchdiffeq
+sentry-sdk==2.54.0
     # via wandb
 shapely==2.1.2
     # via gym-pusht
+shellingham==1.5.4
+    # via typer
 six==1.17.0
     # via
     #   pynput
     #   python-dateutil
-smmap==5.0.2
+smmap==5.0.3
     # via gitdb
-sniffio==1.3.1
-    # via anyio
 stack-data==0.6.3
     # via ipython
-starlette==0.48.0
+starlette==0.52.1
     # via fastapi
 sympy==1.14.0
     # via torch
-teleop==0.1.2
+teleop==0.1.4
     # via lerobot
-tensorboard==2.20.0
-    # via robomimic
-tensorboard-data-server==0.7.2
-    # via tensorboard
-tensorboardx==2.6.4
-    # via robomimic
-termcolor==3.1.0
-    # via
-    #   lerobot
-    #   robomimic
-thop==0.1.1.post2209072238
-    # via libero
-tifffile==2025.5.10
+termcolor==3.3.0
+    # via lerobot
+tifffile==2026.3.3
     # via scikit-image
-timm==1.0.20
-    # via lerobot
-tokenizers==0.22.1
+tokenizers==0.22.2
     # via transformers
 toml==0.10.2
     # via draccus
-tomli==2.3.0
-    # via
-    #   cmeel
-    #   coverage
-    #   jupytext
-    #   pytest
-torch==2.7.1
+torch==2.10.0
     # via
     #   accelerate
     #   lerobot
     #   peft
-    #   robomimic
-    #   thop
-    #   timm
+    #   torchdiffeq
     #   torchvision
-torchcodec==0.5
+torchcodec==0.10.0
     # via lerobot
-torchvision==0.22.1
-    # via
-    #   lerobot
-    #   robomimic
-    #   timm
-tornado==6.5.2
+torchdiffeq==0.2.5
+    # via lerobot
+torchvision==0.25.0
+    # via lerobot
+tornado==6.5.4
     # via meshcat
-tqdm==4.67.1
+tqdm==4.67.3
     # via
     #   datasets
     #   dm-control
     #   huggingface-hub
     #   peft
-    #   robomimic
     #   transformers
 traitlets==5.14.3
     # via
     #   ipython
-    #   jupyter-core
     #   matplotlib-inline
-    #   nbformat
-transformers==4.57.1
+transformers==5.3.0
     # via
     #   lerobot
-    #   libero
     #   peft
 transforms3d==0.4.2
     # via teleop
+typer==0.24.1
+    # via
+    #   huggingface-hub
+    #   transformers
 typing-extensions==4.15.0
     # via
     #   aiosignal
     #   anyio
     #   etils
-    #   exceptiongroup
+    #   faker
     #   fastapi
     #   gymnasium
     #   huggingface-hub
-    #   ipython
-    #   multidict
+    #   mypy
     #   pydantic
     #   pydantic-core
-    #   referencing
     #   rerun-sdk
     #   starlette
     #   torch
     #   typing-inspect
     #   typing-inspection
-    #   uvicorn
-    #   virtualenv
     #   wandb
 typing-inspect==0.9.0
     # via draccus
 typing-inspection==0.4.2
-    # via pydantic
-tzdata==2025.2
+    # via
+    #   fastapi
+    #   pydantic
+tzdata==2025.3
     # via pandas
 u-msgpack-python==2.8.0
     # via meshcat
-urllib3==2.5.0
+urllib3==2.6.3
     # via
     #   requests
     #   sentry-sdk
-uvicorn[standard]==0.38.0
+uvicorn[standard]==0.41.0
     # via teleop
 uvloop==0.22.1
     # via uvicorn
-virtualenv==20.35.3
+virtualenv==21.1.0
     # via pre-commit
-wandb==0.21.4
-    # via
-    #   lerobot
-    #   libero
+wandb==0.24.2
+    # via lerobot
 watchfiles==1.1.1
     # via uvicorn
-wcwidth==0.2.14
+wcwidth==0.6.0
     # via prompt-toolkit
 websocket-client==1.9.0
     # via teleop
-websockets==15.0.1
+websockets==16.0
     # via uvicorn
-werkzeug==3.1.3
-    # via tensorboard
-wrapt==2.0.0
+wrapt==2.1.2
     # via dm-tree
 xxhash==3.6.0
     # via datasets
-yarl==1.22.0
+yarl==1.23.0
     # via aiohttp
 zipp==3.23.0
     # via

requirements-ubuntu.txt

@@ -1,12 +1,12 @@
 #
-# This file is autogenerated by pip-compile with Python 3.10
+# This file is autogenerated by pip-compile with Python 3.12
 # by the following command:
 #
 #    pip-compile --output-file=requirements-ubuntu.txt requirements.in
 #
 -e .[all]
     # via -[all]
-absl-py==2.3.1
+absl-py==2.4.0
     # via
     #   dm-control
     #   dm-env
@@ -14,30 +14,33 @@ absl-py==2.3.1
     #   labmaze
     #   mujoco
     #   tensorboard
-accelerate==1.11.0
+accelerate==1.13.0
     # via
     #   lerobot
     #   peft
 aiohappyeyeballs==2.6.1
     # via aiohttp
-aiohttp==3.13.1
+aiohttp==3.13.3
     # via fsspec
 aiosignal==1.4.0
     # via aiohttp
+annotated-doc==0.0.4
+    # via
+    #   fastapi
+    #   typer
 annotated-types==0.7.0
     # via pydantic
 antlr4-python3-runtime==4.9.3
     # via
     #   hydra-core
     #   omegaconf
-anyio==4.11.0
+anyio==4.12.1
     # via
+    #   httpx
     #   starlette
     #   watchfiles
-asttokens==3.0.0
+asttokens==3.0.1
     # via stack-data
-async-timeout==5.0.1
-    # via aiohttp
 attrs==25.4.0
     # via
     #   aiohttp
@@ -47,30 +50,35 @@ attrs==25.4.0
     #   referencing
     #   rerun-sdk
 av==15.1.0
-    # via lerobot
-bddl==1.0.1
-    # via libero
-certifi==2025.10.5
     # via
+    #   lerobot
+    #   qwen-vl-utils
+bddl==1.0.1
+    # via hf-libero
+certifi==2026.2.25
+    # via
+    #   httpcore
+    #   httpx
     #   requests
     #   sentry-sdk
 cffi==2.0.0
     # via pymunk
-cfgv==3.4.0
+cfgv==3.5.0
     # via pre-commit
-charset-normalizer==3.4.4
+charset-normalizer==3.4.5
     # via requests
-click==8.3.0
+click==8.3.1
     # via
+    #   typer
     #   uvicorn
     #   wandb
-cloudpickle==3.1.1
+cloudpickle==3.1.2
     # via
     #   gymnasium
-    #   libero
-cmake==4.1.0
+    #   hf-libero
+cmake==4.1.3
     # via lerobot
-cmeel==0.57.3
+cmeel==0.59.0
     # via
     #   cmeel-assimp
     #   cmeel-boost
@@ -108,20 +116,24 @@ cmeel-zlib==1.3.1
     # via cmeel-assimp
 coal-library==3.0.1
     # via pin
-contourpy==1.3.2
-    # via matplotlib
-coverage[toml]==7.11.0
+contourpy==1.3.3
+    # via
+    #   lerobot
+    #   matplotlib
+coverage[toml]==7.13.4
     # via pytest-cov
+cuda-bindings==12.9.4
+    # via torch
+cuda-pathfinder==1.4.1
+    # via cuda-bindings
 cycler==0.12.1
     # via matplotlib
-datasets==4.1.1
+datasets==4.6.1
     # via lerobot
-debugpy==1.8.17
+debugpy==1.8.20
     # via lerobot
 decorator==5.2.1
     # via ipython
-decord==0.6.0
-    # via lerobot
 deepdiff==8.6.1
     # via lerobot
 diffusers==0.35.2
@@ -132,7 +144,7 @@ dill==0.4.0
     #   multiprocess
 distlib==0.4.0
     # via virtualenv
-dm-control==1.0.34
+dm-control==1.0.37
     # via gym-aloha
 dm-env==1.6
     # via dm-control
@@ -140,7 +152,6 @@ dm-tree==0.1.9
     # via
     #   dm-control
     #   dm-env
-    #   lerobot
 docopt==0.6.2
     # via num2words
 draccus==0.10.0
@@ -148,66 +159,60 @@ draccus==0.10.0
 dynamixel-sdk==3.8.4
     # via lerobot
 easydict==1.13
-    # via libero
-egl-probe @ git+https://github.com/huggingface/egl_probe.git
-    # via
-    #   libero
-    #   robomimic
+    # via hf-libero
+egl-probe==1.0.2
+    # via robomimic
 eigenpy==3.10.3
     # via coal-library
-einops==0.8.1
+einops==0.8.2
     # via
-    #   flash-attn
+    #   hf-libero
     #   lerobot
-    #   libero
 eiquadprog==1.2.9
     # via placo
-etils[epath,epy]==1.13.0
+etils[epath,epy]==1.14.0
     # via mujoco
-evdev==1.9.2
+evdev==1.9.3
     # via pynput
-exceptiongroup==1.3.0
-    # via
-    #   anyio
-    #   ipython
-    #   pytest
 executing==2.2.1
     # via stack-data
+faker==34.0.2
+    # via lerobot
 farama-notifications==0.0.4
     # via gymnasium
-fastapi==0.119.1
-    # via teleop
+fastapi==0.135.1
+    # via
+    #   lerobot
+    #   teleop
 fastjsonschema==2.21.2
     # via nbformat
 feetech-servo-sdk==1.0.0
     # via lerobot
-filelock==3.20.0
+filelock==3.25.0
     # via
     #   datasets
     #   diffusers
     #   huggingface-hub
+    #   python-discovery
     #   torch
-    #   transformers
     #   virtualenv
-flash-attn==2.8.3
-    # via lerobot
-fonttools==4.60.1
+fonttools==4.61.1
     # via matplotlib
 frozenlist==1.8.0
     # via
     #   aiohttp
     #   aiosignal
-fsspec[http]==2025.9.0
+fsspec[http]==2026.2.0
     # via
     #   datasets
     #   etils
     #   huggingface-hub
     #   torch
 future==1.0.0
-    # via libero
+    # via hf-libero
 gitdb==4.0.12
     # via gitpython
-gitpython==3.1.45
+gitpython==3.1.46
     # via wandb
 glfw==2.10.0
     # via
@@ -230,50 +235,60 @@ gym-hil==0.1.13
     # via lerobot
 gym-pusht==0.1.6
     # via lerobot
-gymnasium==1.2.1
+gymnasium==1.2.3
     # via
     #   gym-aloha
     #   gym-hil
     #   gym-pusht
+    #   hf-libero
     #   lerobot
-    #   libero
     #   metaworld
 h11==0.16.0
-    # via uvicorn
-h5py==3.15.1
+    # via
+    #   httpcore
+    #   uvicorn
+h5py==3.16.0
     # via robomimic
 hebi-py==2.11.0
     # via lerobot
-hf-transfer==0.1.9
-    # via huggingface-hub
-hf-xet==1.1.10
+hf-egl-probe==1.0.2
+    # via hf-libero
+hf-libero==0.1.3
+    # via lerobot
+hf-xet==1.3.2
     # via huggingface-hub
 hidapi==0.14.0.post4
     # via
     #   gym-hil
     #   lerobot
+httpcore==1.0.9
+    # via httpx
 httptools==0.7.1
     # via uvicorn
-huggingface-hub[cli,hf-transfer]==0.35.3
+httpx==0.28.1
+    # via
+    #   datasets
+    #   huggingface-hub
+huggingface-hub==1.6.0
     # via
     #   accelerate
     #   datasets
     #   diffusers
     #   lerobot
     #   peft
-    #   timm
     #   tokenizers
     #   transformers
 hydra-core==1.3.2
-    # via libero
-identify==2.6.15
+    # via hf-libero
+identify==2.6.17
     # via pre-commit
 idna==3.11
     # via
     #   anyio
+    #   httpx
     #   requests
     #   yarl
-imageio[ffmpeg]==2.37.0
+imageio[ffmpeg]==2.37.2
     # via
     #   gym-aloha
     #   gym-hil
@@ -285,16 +300,14 @@ imageio-ffmpeg==0.6.0
     # via
     #   imageio
     #   robomimic
-importlib-metadata==8.7.0
+importlib-metadata==8.7.1
     # via diffusers
-importlib-resources==6.5.2
-    # via etils
 iniconfig==2.3.0
     # via pytest
-inquirerpy==0.3.4
-    # via huggingface-hub
-ipython==8.37.0
+ipython==9.11.0
     # via meshcat
+ipython-pygments-lexers==1.1.1
+    # via ipython
 ischedule==1.2.7
     # via placo
 jedi==0.19.2
@@ -303,40 +316,41 @@ jinja2==3.1.6
     # via torch
 jsonlines==4.0.0
     # via lerobot
-jsonschema==4.25.1
+jsonschema==4.26.0
     # via nbformat
 jsonschema-specifications==2025.9.1
     # via jsonschema
 jupyter-core==5.9.1
     # via nbformat
-jupytext==1.18.1
+jupytext==1.19.1
     # via bddl
 kiwisolver==1.4.9
     # via matplotlib
 labmaze==1.0.6
     # via dm-control
-lazy-loader==0.4
+lazy-loader==0.5
     # via scikit-image
-libero @ git+https://github.com/huggingface/lerobot-libero.git@main
-    # via lerobot
-llvmlite==0.45.1
+librt==0.8.1
+    # via mypy
+llvmlite==0.46.0
     # via numba
 lxml==6.0.2
     # via dm-control
-markdown==3.9
+markdown==3.10.2
     # via tensorboard
 markdown-it-py==4.0.0
     # via
     #   jupytext
     #   mdit-py-plugins
+    #   rich
 markupsafe==3.0.3
     # via
     #   jinja2
     #   werkzeug
-matplotlib==3.10.7
+matplotlib==3.10.8
     # via
+    #   hf-libero
     #   lerobot
-    #   libero
 matplotlib-inline==0.2.1
     # via ipython
 mdit-py-plugins==0.5.0
@@ -353,36 +367,38 @@ mock-serial==0.0.1
     # via lerobot
 mpmath==1.3.0
     # via sympy
-mujoco==3.3.7
+mujoco==3.5.0
     # via
     #   dm-control
     #   gym-aloha
     #   gym-hil
-    #   libero
+    #   hf-libero
     #   metaworld
     #   robosuite
-multidict==6.7.0
+multidict==6.7.1
     # via
     #   aiohttp
     #   yarl
-multiprocess==0.70.16
+multiprocess==0.70.18
     # via datasets
+mypy==1.19.1
+    # via lerobot
 mypy-extensions==1.1.0
-    # via typing-inspect
+    # via
+    #   mypy
+    #   typing-inspect
 nbformat==5.10.4
     # via jupytext
-networkx==3.4.2
+networkx==3.6.1
     # via
     #   bddl
     #   scikit-image
     #   torch
-ninja==1.13.0
-    # via lerobot
-nodeenv==1.9.1
+nodeenv==1.10.0
     # via pre-commit
 num2words==0.5.14
     # via lerobot
-numba==0.62.1
+numba==0.64.0
     # via robosuite
 numpy==2.2.6
     # via
@@ -391,7 +407,6 @@ numpy==2.2.6
     #   cmeel-boost
     #   contourpy
     #   datasets
-    #   decord
     #   diffusers
     #   dm-control
     #   dm-env
@@ -399,9 +414,10 @@ numpy==2.2.6
     #   gymnasium
     #   h5py
     #   hebi-py
+    #   hf-libero
     #   imageio
     #   labmaze
-    #   libero
+    #   lerobot
     #   matplotlib
     #   meshcat
     #   metaworld
@@ -426,49 +442,51 @@ numpy==2.2.6
     #   torchvision
     #   transformers
     #   transforms3d
-nvidia-cublas-cu12==12.6.4.1
+nvidia-cublas-cu12==12.8.4.1
     # via
     #   nvidia-cudnn-cu12
     #   nvidia-cusolver-cu12
     #   torch
-nvidia-cuda-cupti-cu12==12.6.80
+nvidia-cuda-cupti-cu12==12.8.90
     # via torch
-nvidia-cuda-nvrtc-cu12==12.6.77
+nvidia-cuda-nvrtc-cu12==12.8.93
     # via torch
-nvidia-cuda-runtime-cu12==12.6.77
+nvidia-cuda-runtime-cu12==12.8.90
     # via torch
-nvidia-cudnn-cu12==9.5.1.17
+nvidia-cudnn-cu12==9.10.2.21
     # via torch
-nvidia-cufft-cu12==11.3.0.4
+nvidia-cufft-cu12==11.3.3.83
     # via torch
-nvidia-cufile-cu12==1.11.1.6
+nvidia-cufile-cu12==1.13.1.3
     # via torch
-nvidia-curand-cu12==10.3.7.77
+nvidia-curand-cu12==10.3.9.90
     # via torch
-nvidia-cusolver-cu12==11.7.1.2
+nvidia-cusolver-cu12==11.7.3.90
     # via torch
-nvidia-cusparse-cu12==12.5.4.2
+nvidia-cusparse-cu12==12.5.8.93
     # via
     #   nvidia-cusolver-cu12
     #   torch
-nvidia-cusparselt-cu12==0.6.3
+nvidia-cusparselt-cu12==0.7.1
     # via torch
-nvidia-nccl-cu12==2.26.2
+nvidia-nccl-cu12==2.27.5
     # via torch
-nvidia-nvjitlink-cu12==12.6.85
+nvidia-nvjitlink-cu12==12.8.93
     # via
     #   nvidia-cufft-cu12
     #   nvidia-cusolver-cu12
     #   nvidia-cusparse-cu12
     #   torch
-nvidia-nvtx-cu12==12.6.77
+nvidia-nvshmem-cu12==3.4.5
+    # via torch
+nvidia-nvtx-cu12==12.8.90
     # via torch
 omegaconf==2.3.0
     # via hydra-core
-opencv-python==4.12.0.88
+opencv-python==4.13.0.92
     # via
     #   gym-pusht
-    #   libero
+    #   hf-libero
     #   reachy2-sdk
     #   robosuite
 opencv-python-headless==4.12.0.88
@@ -487,6 +505,7 @@ packaging==25.0
     #   matplotlib
     #   peft
     #   pytest
+    #   qwen-vl-utils
     #   reachy2-sdk
     #   scikit-image
     #   tensorboard
@@ -497,21 +516,21 @@ pandas==2.3.3
     # via
     #   datasets
     #   lerobot
-parso==0.8.5
+parso==0.8.6
     # via jedi
-peft==0.17.1
+pathspec==1.0.4
+    # via mypy
+peft==0.18.1
     # via lerobot
 pexpect==4.9.0
     # via ipython
-pfzy==0.3.4
-    # via inquirerpy
-pillow==12.0.0
+pillow==12.1.1
     # via
     #   diffusers
     #   imageio
-    #   lerobot
     #   matplotlib
     #   meshcat
+    #   qwen-vl-utils
     #   rerun-sdk
     #   robosuite
     #   scikit-image
@@ -519,28 +538,27 @@ pillow==12.0.0
     #   torchvision
 pin==3.4.0
     # via placo
-placo==0.9.14
+placo==0.9.16
     # via lerobot
-platformdirs==4.5.0
+platformdirs==4.9.4
     # via
     #   jupyter-core
+    #   python-discovery
     #   virtualenv
     #   wandb
 pluggy==1.6.0
     # via
     #   pytest
     #   pytest-cov
-pre-commit==4.3.0
+pre-commit==4.5.1
     # via lerobot
 prompt-toolkit==3.0.52
-    # via
-    #   inquirerpy
-    #   ipython
+    # via ipython
 propcache==0.4.1
     # via
     #   aiohttp
     #   yarl
-protobuf==6.31.0
+protobuf==6.31.1
     # via
     #   dm-control
     #   grpcio-tools
@@ -550,7 +568,7 @@ protobuf==6.31.0
     #   tensorboard
     #   tensorboardx
     #   wandb
-psutil==7.1.1
+psutil==7.2.2
     # via
     #   accelerate
     #   imageio
@@ -560,17 +578,17 @@ ptyprocess==0.7.0
     # via pexpect
 pure-eval==0.2.3
     # via stack-data
-pyarrow==21.0.0
+pyarrow==23.0.1
     # via
     #   datasets
     #   rerun-sdk
-pycparser==2.23
+pycparser==3.0
     # via cffi
-pydantic==2.12.3
+pydantic==2.12.5
     # via
     #   fastapi
     #   wandb
-pydantic-core==2.41.4
+pydantic-core==2.41.5
     # via pydantic
 pygame==2.6.1
     # via
@@ -580,12 +598,14 @@ pygame==2.6.1
 pygments==2.19.2
     # via
     #   ipython
+    #   ipython-pygments-lexers
     #   pytest
+    #   rich
 pymunk==6.11.1
     # via
     #   gym-pusht
     #   lerobot
-pyngrok==7.4.1
+pyngrok==7.5.1
     # via meshcat
 pynput==1.8.1
     # via
@@ -595,7 +615,7 @@ pyopengl==3.1.10
     # via
     #   dm-control
     #   mujoco
-pyparsing==3.2.5
+pyparsing==3.3.2
     # via
     #   dm-control
     #   matplotlib
@@ -621,13 +641,16 @@ pytest-timeout==2.4.0
     # via lerobot
 python-dateutil==2.9.0.post0
     # via
+    #   faker
     #   matplotlib
     #   pandas
-python-dotenv==1.1.1
+python-discovery==1.1.1
+    # via virtualenv
+python-dotenv==1.2.2
     # via uvicorn
 python-xlib==0.33
     # via pynput
-pytz==2025.2
+pytz==2026.1.post1
     # via pandas
 pyyaml==6.0.3
     # via
@@ -642,7 +665,6 @@ pyyaml==6.0.3
     #   pre-commit
     #   pyngrok
     #   pyyaml-include
-    #   timm
     #   transformers
     #   uvicorn
     #   wandb
@@ -652,7 +674,9 @@ pyzmq==27.1.0
     # via
     #   lerobot
     #   meshcat
-reachy2-sdk==1.0.14
+qwen-vl-utils==0.0.14
+    # via lerobot
+reachy2-sdk==1.0.15
     # via lerobot
 reachy2-sdk-api==1.0.21
     # via reachy2-sdk
@@ -660,7 +684,7 @@ referencing==0.37.0
     # via
     #   jsonschema
     #   jsonschema-specifications
-regex==2025.10.23
+regex==2026.2.28
     # via
     #   diffusers
     #   transformers
@@ -669,60 +693,62 @@ requests==2.32.5
     #   datasets
     #   diffusers
     #   dm-control
-    #   huggingface-hub
+    #   qwen-vl-utils
     #   teleop
-    #   transformers
     #   wandb
-rerun-sdk==0.26.1
+rerun-sdk==0.26.2
     # via lerobot
 rhoban-cmeel-jsoncpp==1.9.4.9
     # via placo
+rich==14.3.3
+    # via typer
 robomimic==0.2.0
-    # via libero
+    # via hf-libero
 robosuite==1.4.0
-    # via libero
-rpds-py==0.28.0
+    # via hf-libero
+rpds-py==0.30.0
     # via
     #   jsonschema
     #   referencing
-safetensors==0.6.2
+safetensors==0.7.0
     # via
     #   accelerate
     #   diffusers
     #   lerobot
     #   peft
-    #   timm
     #   transformers
 scikit-image==0.25.2
     # via
     #   gym-pusht
     #   lerobot
-scipy==1.15.3
+scipy==1.17.1
     # via
     #   dm-control
+    #   lerobot
     #   metaworld
     #   robosuite
     #   scikit-image
-sentry-sdk==2.42.1
+    #   torchdiffeq
+sentry-sdk==2.54.0
     # via wandb
 shapely==2.1.2
     # via gym-pusht
+shellingham==1.5.4
+    # via typer
 six==1.17.0
     # via
     #   pynput
     #   python-dateutil
     #   python-xlib
-smmap==5.0.2
+smmap==5.0.3
     # via gitdb
-sniffio==1.3.1
-    # via anyio
 stack-data==0.6.3
     # via ipython
-starlette==0.48.0
+starlette==0.52.1
     # via fastapi
 sympy==1.14.0
     # via torch
-teleop==0.1.2
+teleop==0.1.4
     # via lerobot
 tensorboard==2.20.0
     # via robomimic
@@ -730,46 +756,38 @@ tensorboard-data-server==0.7.2
     # via tensorboard
 tensorboardx==2.6.4
     # via robomimic
-termcolor==3.1.0
+termcolor==3.3.0
     # via
     #   lerobot
     #   robomimic
 thop==0.1.1.post2209072238
-    # via libero
-tifffile==2025.5.10
+    # via hf-libero
+tifffile==2026.3.3
     # via scikit-image
-timm==1.0.20
-    # via lerobot
-tokenizers==0.22.1
+tokenizers==0.22.2
     # via transformers
 toml==0.10.2
     # via draccus
-tomli==2.3.0
-    # via
-    #   cmeel
-    #   coverage
-    #   jupytext
-    #   pytest
-torch==2.7.1
+torch==2.10.0
     # via
     #   accelerate
-    #   flash-attn
     #   lerobot
     #   peft
     #   robomimic
     #   thop
-    #   timm
+    #   torchdiffeq
     #   torchvision
-torchcodec==0.5
+torchcodec==0.10.0
     # via lerobot
-torchvision==0.22.1
+torchdiffeq==0.2.5
+    # via lerobot
+torchvision==0.25.0
     # via
     #   lerobot
     #   robomimic
-    #   timm
-tornado==6.5.2
+tornado==6.5.4
     # via meshcat
-tqdm==4.67.1
+tqdm==4.67.3
     # via
     #   datasets
     #   dm-control
@@ -783,26 +801,29 @@ traitlets==5.14.3
     #   jupyter-core
     #   matplotlib-inline
     #   nbformat
-transformers==4.57.1
+transformers==5.3.0
     # via
+    #   hf-libero
     #   lerobot
-    #   libero
     #   peft
 transforms3d==0.4.2
     # via teleop
-triton==3.3.1
+triton==3.6.0
     # via torch
+typer==0.24.1
+    # via
+    #   huggingface-hub
+    #   transformers
 typing-extensions==4.15.0
     # via
     #   aiosignal
     #   anyio
     #   etils
-    #   exceptiongroup
+    #   faker
     #   fastapi
     #   gymnasium
     #   huggingface-hub
-    #   ipython
-    #   multidict
+    #   mypy
     #   pydantic
     #   pydantic-core
     #   referencing
@@ -811,46 +832,46 @@ typing-extensions==4.15.0
     #   torch
     #   typing-inspect
     #   typing-inspection
-    #   uvicorn
-    #   virtualenv
     #   wandb
 typing-inspect==0.9.0
     # via draccus
 typing-inspection==0.4.2
-    # via pydantic
-tzdata==2025.2
+    # via
+    #   fastapi
+    #   pydantic
+tzdata==2025.3
     # via pandas
 u-msgpack-python==2.8.0
     # via meshcat
-urllib3==2.5.0
+urllib3==2.6.3
     # via
     #   requests
     #   sentry-sdk
-uvicorn[standard]==0.38.0
+uvicorn[standard]==0.41.0
     # via teleop
 uvloop==0.22.1
     # via uvicorn
-virtualenv==20.35.3
+virtualenv==21.1.0
     # via pre-commit
-wandb==0.21.4
+wandb==0.24.2
     # via
+    #   hf-libero
     #   lerobot
-    #   libero
 watchfiles==1.1.1
     # via uvicorn
-wcwidth==0.2.14
+wcwidth==0.6.0
     # via prompt-toolkit
 websocket-client==1.9.0
     # via teleop
-websockets==15.0.1
+websockets==16.0
     # via uvicorn
-werkzeug==3.1.3
+werkzeug==3.1.6
     # via tensorboard
-wrapt==2.0.0
+wrapt==2.1.2
     # via dm-tree
 xxhash==3.6.0
     # via datasets
-yarl==1.22.0
+yarl==1.23.0
     # via aiohttp
 zipp==3.23.0
     # via

requirements.in

@@ -1,9 +1,9 @@
 # requirements.in
 
-# requirements-macos.txt was generated on macOS and is platform-specific (macOS 26.0.1 25A362 arm64).
-# Darwin MacBook-Pro.local 25.0.0 Darwin Kernel Version 25.0.0: Wed Sep 17 21:42:08 PDT 2025; root:xnu-12377.1.9~141/RELEASE_ARM64_T8132 arm64
+# requirements-macos.txt was generated on macOS and is platform-specific (macOS 26.3.1 25D2128 arm64).
+# Darwin MacBook-Pro.local 25.3.0 Darwin Kernel Version 25.3.0: Wed Jan 28 20:54:55 PST 2026; root:xnu-12377.91.3~2/RELEASE_ARM64_T8132 arm64
 
-# requirements-ubuntu.txt was generated on Linux and is platform-specific (Ubuntu 24.04.3 LTS x86_64).
-# Linux mlerobot-linux 6.14.0-33-generic #33~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Fri Sep 19 17:02:30 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
+# requirements-ubuntu.txt was generated on Linux and is platform-specific (Ubuntu 24.04.4 LTS x86_64).
+# Linux lerobot-linux 6.17.0-14-generic #14~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Jan 15 15:52:10 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
 
 -e .[all]

src/lerobot/__init__.py

@@ -29,6 +29,7 @@ Example:
         print(lerobot.available_policies_per_env)
         print(lerobot.available_robots)
         print(lerobot.available_cameras)
+        print(lerobot.available_microphones)
         print(lerobot.available_motors)
     ```
 
@@ -174,6 +175,12 @@ available_cameras = [
     "intelrealsense",
 ]
 
+# lists all available microphones from `lerobot/microphones`
+available_microphones = [
+    "portaudio",
+    "touchlab",
+]
+
 # lists all available motors from `lerobot/motors`
 available_motors = [
     "dynamixel",

src/lerobot/async_inference/helpers.py

@@ -23,7 +23,7 @@ from typing import Any
 import torch
 
 from lerobot.configs.types import PolicyFeature
-from lerobot.datasets.utils import build_dataset_frame, hw_to_dataset_features
+from lerobot.datasets.feature_utils import build_dataset_frame, hw_to_dataset_features
 
 # NOTE: Configs need to be loaded for the client to be able to instantiate the policy config
 from lerobot.policies import (  # noqa: F401

src/lerobot/async_inference/policy_server.py

@@ -39,15 +39,13 @@ import grpc
 import torch
 
 from lerobot.policies.factory import get_policy_class, make_pre_post_processors
-from lerobot.processor import (
-    PolicyAction,
-    PolicyProcessorPipeline,
-)
+from lerobot.processor import PolicyProcessorPipeline
 from lerobot.transport import (
     services_pb2,  # type: ignore
     services_pb2_grpc,  # type: ignore
 )
 from lerobot.transport.utils import receive_bytes_in_chunks
+from lerobot.types import PolicyAction
 
 from .configs import PolicyServerConfig
 from .constants import SUPPORTED_POLICIES

src/lerobot/async_inference/robot_client.py

@@ -49,6 +49,8 @@ import torch
 
 from lerobot.cameras.opencv.configuration_opencv import OpenCVCameraConfig  # noqa: F401
 from lerobot.cameras.realsense.configuration_realsense import RealSenseCameraConfig  # noqa: F401
+from lerobot.microphones.portaudio.configuration_portaudio import PortAudioMicrophoneConfig  # noqa: F401
+from lerobot.microphones.touchlab.configuration_touchlab import TouchLabSensorConfig  # noqa: F401
 from lerobot.robots import (  # noqa: F401
     Robot,
     RobotConfig,
@@ -63,9 +65,9 @@ from lerobot.transport import (
     services_pb2_grpc,  # type: ignore
 )
 from lerobot.transport.utils import grpc_channel_options, send_bytes_in_chunks
+from lerobot.utils.import_utils import register_third_party_plugins
 
 from .configs import RobotClientConfig
-from .constants import SUPPORTED_ROBOTS
 from .helpers import (
     Action,
     FPSTracker,
@@ -485,8 +487,9 @@ class RobotClient:
 def async_client(cfg: RobotClientConfig):
     logging.info(pformat(asdict(cfg)))
 
-    if cfg.robot.type not in SUPPORTED_ROBOTS:
-        raise ValueError(f"Robot {cfg.robot.type} not yet supported!")
+    # TODO: Assert if checking robot support is still needed with the plugin system
+    # if cfg.robot.type not in SUPPORTED_ROBOTS:
+    #     raise ValueError(f"Robot {cfg.robot.type} not yet supported!")
 
     client = RobotClient(cfg)
 
@@ -512,4 +515,5 @@ def async_client(cfg: RobotClientConfig):
 
 
 if __name__ == "__main__":
+    register_third_party_plugins()
     async_client()  # run the client

src/lerobot/cameras/camera.py

@@ -150,7 +150,7 @@ class Camera(abc.ABC):
         """
         pass
 
-    def read_latest(self, max_age_ms: int = 1000) -> NDArray[Any]:
+    def read_latest(self, max_age_ms: int = 500) -> NDArray[Any]:
         """Return the most recent frame captured immediately (Peeking).
 
         This method is non-blocking and returns whatever is currently in the

src/lerobot/cameras/opencv/camera_opencv.py

@@ -530,7 +530,7 @@ class OpenCVCamera(Camera):
         return frame
 
     @check_if_not_connected
-    def read_latest(self, max_age_ms: int = 1000) -> NDArray[Any]:
+    def read_latest(self, max_age_ms: int = 500) -> NDArray[Any]:
         """Return the most recent frame captured immediately (Peeking).
 
         This method is non-blocking and returns whatever is currently in the

src/lerobot/cameras/reachy2_camera/reachy2_camera.py

@@ -201,7 +201,7 @@ class Reachy2Camera(Camera):
         return self.read()
 
     @check_if_not_connected
-    def read_latest(self, max_age_ms: int = 1000) -> NDArray[Any]:
+    def read_latest(self, max_age_ms: int = 500) -> NDArray[Any]:
         """Return the most recent frame captured immediately (Peeking).
 
         This method is non-blocking and returns whatever is currently in the

src/lerobot/cameras/realsense/camera_realsense.py

@@ -573,7 +573,7 @@ class RealSenseCamera(Camera):
 
     # NOTE(Steven): Missing implementation for depth for now
     @check_if_not_connected
-    def read_latest(self, max_age_ms: int = 1000) -> NDArray[Any]:
+    def read_latest(self, max_age_ms: int = 500) -> NDArray[Any]:
         """Return the most recent (color) frame captured immediately (Peeking).
 
         This method is non-blocking and returns whatever is currently in the

src/lerobot/cameras/zmq/camera_zmq.py

@@ -181,7 +181,7 @@ class ZMQCamera(Camera):
         try:
             message = self.socket.recv_string()
         except Exception as e:
-            # Check for ZMQ timeout (EAGAIN/Again) without requiring global zmq import
+            # zmq is lazy-imported in connect(), so check by name to avoid a top-level import
             if type(e).__name__ == "Again":
                 raise TimeoutError(f"{self} timeout after {self.timeout_ms}ms") from e
             raise

src/lerobot/cameras/zmq/image_server.py

@@ -23,6 +23,7 @@ import base64
 import contextlib
 import json
 import logging
+import threading
 import time
 from collections import deque
 
@@ -42,10 +43,57 @@ def encode_image(image: np.ndarray, quality: int = 80) -> str:
     return base64.b64encode(buffer).decode("utf-8")
 
 
+class CameraCaptureThread:
+    """Background thread that continuously captures and encodes frames from a camera."""
+
+    def __init__(self, camera: OpenCVCamera, name: str):
+        self.camera = camera
+        self.name = name
+        self.latest_encoded: str | None = None  # Pre-encoded JPEG as base64
+        self.latest_timestamp: float = 0.0
+        self.frame_lock = threading.Lock()
+        self.running = False
+        self.thread: threading.Thread | None = None
+
+    def start(self):
+        """Start the capture thread."""
+        self.running = True
+        self.thread = threading.Thread(target=self._capture_loop, daemon=True)
+        self.thread.start()
+
+    def stop(self):
+        """Stop the capture thread."""
+        self.running = False
+        if self.thread:
+            self.thread.join(timeout=1.0)
+
+    def _capture_loop(self):
+        """Continuously capture and encode frames at the camera's native rate."""
+        while self.running:
+            try:
+                frame = self.camera.read()  # Blocks at camera's native rate
+                timestamp = time.time()
+                # Encode immediately in capture thread (this is the slow part)
+                encoded = encode_image(frame)
+                with self.frame_lock:
+                    self.latest_encoded = encoded
+                    self.latest_timestamp = timestamp
+            except Exception as e:
+                logger.warning(f"Camera {self.name} capture error: {e}")
+                time.sleep(0.01)
+
+    def get_latest(self) -> tuple[str | None, float]:
+        """Get the latest encoded frame and its timestamp."""
+        with self.frame_lock:
+            return self.latest_encoded, self.latest_timestamp
+
+
 class ImageServer:
     def __init__(self, config: dict, port: int = 5555):
+        # fps controls the publish loop rate (how often frames are sent over ZMQ), not the camera capture rate
         self.fps = config.get("fps", 30)
         self.cameras: dict[str, OpenCVCamera] = {}
+        self.capture_threads: dict[str, CameraCaptureThread] = {}
 
         for name, cfg in config.get("cameras", {}).items():
             shape = cfg.get("shape", [480, 640])
@@ -61,6 +109,10 @@ class ImageServer:
             self.cameras[name] = camera
             logger.info(f"Camera {name}: {shape[1]}x{shape[0]}")
 
+            # Create capture thread for this camera
+            capture_thread = CameraCaptureThread(camera, name)
+            self.capture_threads[name] = capture_thread
+
         # ZMQ PUB socket
         self.context = zmq.Context()
         self.socket = self.context.socket(zmq.PUB)
@@ -73,6 +125,18 @@ class ImageServer:
     def run(self):
         frame_count = 0
         frame_times = deque(maxlen=60)
+        last_published_ts: dict[str, float] = {}
+
+        # Start all capture threads
+        for capture_thread in self.capture_threads.values():
+            capture_thread.start()
+
+        # Wait for first frames to be captured and encoded
+        logger.info("Waiting for cameras to start capturing...")
+        for name, capture_thread in self.capture_threads.items():
+            while capture_thread.get_latest()[0] is None:
+                time.sleep(0.01)
+            logger.info(f"Camera {name} ready (capture + encode in background)")
 
         try:
             while True:
@@ -80,10 +144,12 @@ class ImageServer:
 
                 # Build message
                 message = {"timestamps": {}, "images": {}}
-                for name, cam in self.cameras.items():
-                    frame = cam.read()  # Returns RGB
-                    message["timestamps"][name] = time.time()
-                    message["images"][name] = encode_image(frame)
+                for name, capture_thread in self.capture_threads.items():
+                    encoded, timestamp = capture_thread.get_latest()
+                    if encoded is not None and timestamp > last_published_ts.get(name, 0.0):
+                        message["timestamps"][name] = timestamp
+                        message["images"][name] = encoded
+                        last_published_ts[name] = timestamp
 
                 # Send as JSON string (suppress if buffer full)
                 with contextlib.suppress(zmq.Again):
@@ -102,6 +168,8 @@ class ImageServer:
         except KeyboardInterrupt:
             pass
         finally:
+            for capture_thread in self.capture_threads.values():
+                capture_thread.stop()
             for cam in self.cameras.values():
                 cam.disconnect()
             self.socket.close()

src/lerobot/configs/default.py

@@ -27,7 +27,8 @@ class DatasetConfig:
     # "dataset_index" into the returned item. The index mapping is made according to the order in which the
     # datasets are provided.
     repo_id: str
-    # Root directory where the dataset will be stored (e.g. 'dataset/path').
+    # Root directory for a concrete local dataset tree (e.g. 'dataset/path'). If None, local datasets are
+    # looked up under $HF_LEROBOT_HOME/repo_id and Hub downloads use a revision-safe cache under $HF_LEROBOT_HOME/hub.
     root: str | None = None
     episodes: list[int] | None = None
     image_transforms: ImageTransformsConfig = field(default_factory=ImageTransformsConfig)
@@ -36,6 +37,16 @@ class DatasetConfig:
     video_backend: str = field(default_factory=get_safe_default_codec)
     streaming: bool = False
 
+    def __post_init__(self) -> None:
+        if self.episodes is not None:
+            if any(ep < 0 for ep in self.episodes):
+                raise ValueError(
+                    f"Episode indices must be non-negative, got: {[ep for ep in self.episodes if ep < 0]}"
+                )
+            if len(self.episodes) != len(set(self.episodes)):
+                duplicates = sorted({ep for ep in self.episodes if self.episodes.count(ep) > 1})
+                raise ValueError(f"Episode indices contain duplicates: {duplicates}")
+
 
 @dataclass
 class WandBConfig:
@@ -47,6 +58,7 @@ class WandBConfig:
     notes: str | None = None
     run_id: str | None = None
     mode: str | None = None  # Allowed values: 'online', 'offline' 'disabled'. Defaults to 'online'
+    add_tags: bool = True  # If True, save configuration as tags in the WandB run.
 
 
 @dataclass

src/lerobot/configs/policies.py

@@ -30,8 +30,8 @@ from lerobot.configs.types import FeatureType, PolicyFeature
 from lerobot.optim.optimizers import OptimizerConfig
 from lerobot.optim.schedulers import LRSchedulerConfig
 from lerobot.utils.constants import ACTION, OBS_STATE
+from lerobot.utils.device_utils import auto_select_torch_device, is_amp_available, is_torch_device_available
 from lerobot.utils.hub import HubMixin
-from lerobot.utils.utils import auto_select_torch_device, is_amp_available, is_torch_device_available
 
 T = TypeVar("T", bound="PreTrainedConfig")
 logger = getLogger(__name__)
@@ -151,6 +151,12 @@ class PreTrainedConfig(draccus.ChoiceRegistry, HubMixin, abc.ABC):  # type: igno
             return {}
         return {key: ft for key, ft in self.input_features.items() if ft.type is FeatureType.VISUAL}
 
+    @property
+    def audio_features(self) -> dict[str, PolicyFeature]:
+        if not self.input_features:
+            return {}
+        return {key: ft for key, ft in self.input_features.items() if ft.type is FeatureType.AUDIO}
+
     @property
     def action_feature(self) -> PolicyFeature | None:
         if not self.output_features:

src/lerobot/configs/train.py

@@ -50,6 +50,9 @@ class TrainPipelineConfig(HubMixin):
     # `seed` is used for training (eg: model initialization, dataset shuffling)
     # AND for the evaluation environments.
     seed: int | None = 1000
+    # Set to True to use deterministic cuDNN algorithms for reproducibility.
+    # This disables cudnn.benchmark and may reduce training speed by ~10-20 percent.
+    cudnn_deterministic: bool = False
     # Number of workers for the dataloader.
     num_workers: int = 4
     batch_size: int = 8

src/lerobot/configs/types.py

@@ -20,6 +20,7 @@ from enum import Enum
 class FeatureType(str, Enum):
     STATE = "STATE"
     VISUAL = "VISUAL"
+    AUDIO = "AUDIO"
     ENV = "ENV"
     ACTION = "ACTION"
     REWARD = "REWARD"

src/lerobot/data_processing/sarm_annotations/subtask_annotation.py

@@ -746,7 +746,8 @@ def save_annotations_to_dataset(
     dataset_path: Path, annotations: dict[int, SubtaskAnnotation], fps: int, prefix: str = "sparse"
 ):
     """Save annotations to LeRobot dataset parquet format."""
-    from lerobot.datasets.utils import DEFAULT_EPISODES_PATH, load_episodes
+    from lerobot.datasets.io_utils import load_episodes
+    from lerobot.datasets.utils import DEFAULT_EPISODES_PATH
 
     episodes_dataset = load_episodes(dataset_path)
     if not episodes_dataset or len(episodes_dataset) == 0:
@@ -840,7 +841,7 @@ def generate_auto_sparse_annotations(
 
 def load_annotations_from_dataset(dataset_path: Path, prefix: str = "sparse") -> dict[int, SubtaskAnnotation]:
     """Load annotations from LeRobot dataset parquet files."""
-    from lerobot.datasets.utils import load_episodes
+    from lerobot.datasets.io_utils import load_episodes
 
     episodes_dataset = load_episodes(dataset_path)
     if not episodes_dataset or len(episodes_dataset) == 0:

src/lerobot/datasets/__init__.py

@@ -0,0 +1,33 @@
+#!/usr/bin/env python
+
+# Copyright 2026 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 lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.lerobot_dataset import LeRobotDataset
+from lerobot.datasets.multi_dataset import MultiLeRobotDataset
+from lerobot.datasets.sampler import EpisodeAwareSampler
+from lerobot.datasets.streaming_dataset import StreamingLeRobotDataset
+from lerobot.datasets.transforms import ImageTransforms, ImageTransformsConfig
+
+__all__ = [
+    "EpisodeAwareSampler",
+    "ImageTransforms",
+    "ImageTransformsConfig",
+    "LeRobotDataset",
+    "LeRobotDatasetMetadata",
+    "MultiLeRobotDataset",
+    "StreamingLeRobotDataset",
+]

src/lerobot/datasets/aggregate.py

@@ -24,24 +24,28 @@ import pandas as pd
 import tqdm
 
 from lerobot.datasets.compute_stats import aggregate_stats
-from lerobot.datasets.lerobot_dataset import LeRobotDatasetMetadata
+from lerobot.datasets.dataset_metadata import LeRobotDatasetMetadata
+from lerobot.datasets.feature_utils import get_hf_features_from_features
+from lerobot.datasets.io_utils import (
+    get_file_size_in_mb,
+    get_parquet_file_size_in_mb,
+    to_parquet_with_hf_images,
+    write_info,
+    write_stats,
+    write_tasks,
+)
 from lerobot.datasets.utils import (
+    DEFAULT_AUDIO_FILE_SIZE_IN_MB,
+    DEFAULT_AUDIO_PATH,
     DEFAULT_CHUNK_SIZE,
     DEFAULT_DATA_FILE_SIZE_IN_MB,
     DEFAULT_DATA_PATH,
     DEFAULT_EPISODES_PATH,
     DEFAULT_VIDEO_FILE_SIZE_IN_MB,
     DEFAULT_VIDEO_PATH,
-    get_file_size_in_mb,
-    get_hf_features_from_features,
-    get_parquet_file_size_in_mb,
-    to_parquet_with_hf_images,
     update_chunk_file_indices,
-    write_info,
-    write_stats,
-    write_tasks,
 )
-from lerobot.datasets.video_utils import concatenate_video_files, get_video_duration_in_s
+from lerobot.datasets.video_utils import concatenate_media_files, get_media_duration_in_s
 
 
 def validate_all_metadata(all_metadata: list[LeRobotDatasetMetadata]):
@@ -110,6 +114,7 @@ def update_meta_data(
     meta_idx,
     data_idx,
     videos_idx,
+    audios_idx,
 ):
     """Updates metadata DataFrame with new chunk, file, and timestamp indices.
 
@@ -125,7 +130,7 @@ def update_meta_data(
         meta_idx: Dictionary containing current metadata chunk and file indices.
         data_idx: Dictionary containing current data chunk and file indices.
         videos_idx: Dictionary containing current video indices and timestamps.
-
+        audios_idx: Dictionary containing current audio indices and timestamps.
     Returns:
         pd.DataFrame: Updated DataFrame with adjusted indices and timestamps.
     """
@@ -223,6 +228,36 @@ def update_meta_data(
         # Clean up temporary columns
         df = df.drop(columns=["_orig_chunk", "_orig_file"])
 
+    for key, audio_idx in audios_idx.items():
+        # Store original audio file indices before updating
+        orig_chunk_col = f"audio/{key}/chunk_index"
+        orig_file_col = f"audio/{key}/file_index"
+        df["_orig_chunk"] = df[orig_chunk_col].copy()
+        df["_orig_file"] = df[orig_file_col].copy()
+
+        # Update chunk and file indices to point to destination
+        df[orig_chunk_col] = audio_idx["chunk"]
+        df[orig_file_col] = audio_idx["file"]
+
+        # Apply per-source-file timestamp offsets
+        src_to_offset = audio_idx.get("src_to_offset", {})
+        if src_to_offset:
+            # Apply offset based on original source file
+            for idx in df.index:
+                src_key = (df.at[idx, "_orig_chunk"], df.at[idx, "_orig_file"])
+                offset = src_to_offset.get(src_key, 0)
+                df.at[idx, f"audio/{key}/from_timestamp"] += offset
+                df.at[idx, f"audio/{key}/to_timestamp"] += offset
+        else:
+            # Fallback to simple offset (for backward compatibility)
+            df[f"audio/{key}/from_timestamp"] = (
+                df[f"audio/{key}/from_timestamp"] + audio_idx["latest_duration"]
+            )
+            df[f"audio/{key}/to_timestamp"] = df[f"audio/{key}/to_timestamp"] + audio_idx["latest_duration"]
+
+        # Clean up temporary columns
+        df = df.drop(columns=["_orig_chunk", "_orig_file"])
+
     df["dataset_from_index"] = df["dataset_from_index"] + dst_meta.info["total_frames"]
     df["dataset_to_index"] = df["dataset_to_index"] + dst_meta.info["total_frames"]
     df["episode_index"] = df["episode_index"] + dst_meta.info["total_episodes"]
@@ -237,6 +272,7 @@ def aggregate_datasets(
     aggr_root: Path | None = None,
     data_files_size_in_mb: float | None = None,
     video_files_size_in_mb: float | None = None,
+    audio_files_size_in_mb: float | None = None,
     chunk_size: int | None = None,
 ):
     """Aggregates multiple LeRobot datasets into a single unified dataset.
@@ -254,6 +290,7 @@ def aggregate_datasets(
         aggr_root: Optional root path for the aggregated dataset.
         data_files_size_in_mb: Maximum size for data files in MB (defaults to DEFAULT_DATA_FILE_SIZE_IN_MB)
         video_files_size_in_mb: Maximum size for video files in MB (defaults to DEFAULT_VIDEO_FILE_SIZE_IN_MB)
+        audio_files_size_in_mb: Maximum size for audio files in MB (defaults to DEFAULT_AUDIO_FILE_SIZE_IN_MB)
         chunk_size: Maximum number of files per chunk (defaults to DEFAULT_CHUNK_SIZE)
     """
     logging.info("Start aggregate_datasets")
@@ -262,6 +299,8 @@ def aggregate_datasets(
         data_files_size_in_mb = DEFAULT_DATA_FILE_SIZE_IN_MB
     if video_files_size_in_mb is None:
         video_files_size_in_mb = DEFAULT_VIDEO_FILE_SIZE_IN_MB
+    if audio_files_size_in_mb is None:
+        audio_files_size_in_mb = DEFAULT_AUDIO_FILE_SIZE_IN_MB
     if chunk_size is None:
         chunk_size = DEFAULT_CHUNK_SIZE
 
@@ -274,6 +313,7 @@ def aggregate_datasets(
     )
     fps, robot_type, features = validate_all_metadata(all_metadata)
     video_keys = [key for key in features if features[key]["dtype"] == "video"]
+    audio_keys = [key for key in features if features[key]["dtype"] == "audio"]
 
     dst_meta = LeRobotDatasetMetadata.create(
         repo_id=aggr_repo_id,
@@ -285,25 +325,32 @@ def aggregate_datasets(
         chunks_size=chunk_size,
         data_files_size_in_mb=data_files_size_in_mb,
         video_files_size_in_mb=video_files_size_in_mb,
+        audio_files_size_in_mb=audio_files_size_in_mb,
     )
 
     logging.info("Find all tasks")
     unique_tasks = pd.concat([m.tasks for m in all_metadata]).index.unique()
-    dst_meta.tasks = pd.DataFrame({"task_index": range(len(unique_tasks))}, index=unique_tasks)
+    dst_meta.tasks = pd.DataFrame(
+        {"task_index": range(len(unique_tasks))}, index=pd.Index(unique_tasks, name="task")
+    )
 
     meta_idx = {"chunk": 0, "file": 0}
     data_idx = {"chunk": 0, "file": 0}
     videos_idx = {
         key: {"chunk": 0, "file": 0, "latest_duration": 0, "episode_duration": 0} for key in video_keys
     }
+    audios_idx = {
+        key: {"chunk": 0, "file": 0, "latest_duration": 0, "episode_duration": 0} for key in audio_keys
+    }
 
     dst_meta.episodes = {}
 
     for src_meta in tqdm.tqdm(all_metadata, desc="Copy data and videos"):
         videos_idx = aggregate_videos(src_meta, dst_meta, videos_idx, video_files_size_in_mb, chunk_size)
+        audios_idx = aggregate_audio(src_meta, dst_meta, audios_idx, audio_files_size_in_mb, chunk_size)
         data_idx = aggregate_data(src_meta, dst_meta, data_idx, data_files_size_in_mb, chunk_size)
 
-        meta_idx = aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx)
+        meta_idx = aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx, audios_idx)
 
         # Clear the src_to_dst mapping after processing each source dataset
         # to avoid interference between different source datasets
@@ -371,7 +418,7 @@ def aggregate_videos(src_meta, dst_meta, videos_idx, video_files_size_in_mb, chu
                 file_index=file_idx,
             )
 
-            src_duration = get_video_duration_in_s(src_path)
+            src_duration = get_media_duration_in_s(src_path, media_type="video")
             dst_key = (chunk_idx, file_idx)
 
             if not dst_path.exists():
@@ -410,7 +457,7 @@ def aggregate_videos(src_meta, dst_meta, videos_idx, video_files_size_in_mb, chu
                 current_dst_duration = dst_file_durations.get(dst_key, 0)
                 videos_idx[key]["src_to_offset"][(src_chunk_idx, src_file_idx)] = current_dst_duration
                 videos_idx[key]["src_to_dst"][(src_chunk_idx, src_file_idx)] = dst_key
-                concatenate_video_files(
+                concatenate_media_files(
                     [dst_path, src_path],
                     dst_path,
                 )
@@ -425,6 +472,101 @@ def aggregate_videos(src_meta, dst_meta, videos_idx, video_files_size_in_mb, chu
     return videos_idx
 
 
+def aggregate_audio(src_meta, dst_meta, audios_idx, audio_files_size_in_mb, chunk_size):
+    """Aggregates audio files from a source dataset into the destination dataset.
+
+    Handles audio file concatenation and rotation based on file size limits.
+    Creates new audio files when size limits are exceeded.
+
+    Args:
+        src_meta: Source dataset metadata.
+        dst_meta: Destination dataset metadata.
+        audio_idx: Dictionary tracking audio chunk and file indices.
+        audio_files_size_in_mb: Maximum size for audio files in MB (defaults to DEFAULT_AUDIO_FILE_SIZE_IN_MB)
+        chunk_size: Maximum number of files per chunk (defaults to DEFAULT_CHUNK_SIZE)
+
+    Returns:
+        dict: Updated audio_idx with current chunk and file indices.
+    """
+    for key in audios_idx:
+        audios_idx[key]["episode_duration"] = 0
+        # Track offset for each source (chunk, file) pair
+        audios_idx[key]["src_to_offset"] = {}
+
+    for key, audio_idx in audios_idx.items():
+        unique_chunk_file_pairs = {
+            (chunk, file)
+            for chunk, file in zip(
+                src_meta.episodes[f"audio/{key}/chunk_index"],
+                src_meta.episodes[f"audio/{key}/file_index"],
+                strict=False,
+            )
+        }
+        unique_chunk_file_pairs = sorted(unique_chunk_file_pairs)
+
+        chunk_idx = audio_idx["chunk"]
+        file_idx = audio_idx["file"]
+        current_offset = audio_idx["latest_duration"]
+
+        for src_chunk_idx, src_file_idx in unique_chunk_file_pairs:
+            src_path = src_meta.root / DEFAULT_AUDIO_PATH.format(
+                audio_key=key,
+                chunk_index=src_chunk_idx,
+                file_index=src_file_idx,
+            )
+
+            dst_path = dst_meta.root / DEFAULT_AUDIO_PATH.format(
+                audio_key=key,
+                chunk_index=chunk_idx,
+                file_index=file_idx,
+            )
+
+            src_duration = get_media_duration_in_s(src_path, media_type="audio")
+
+            if not dst_path.exists():
+                # Store offset before incrementing
+                audios_idx[key]["src_to_offset"][(src_chunk_idx, src_file_idx)] = current_offset
+                dst_path.parent.mkdir(parents=True, exist_ok=True)
+                shutil.copy(str(src_path), str(dst_path))
+                audios_idx[key]["episode_duration"] += src_duration
+                current_offset += src_duration
+                continue
+
+            # Check file sizes before appending
+            src_size = get_file_size_in_mb(src_path)
+            dst_size = get_file_size_in_mb(dst_path)
+
+            if dst_size + src_size >= audio_files_size_in_mb:
+                # Rotate to a new file, this source becomes start of new destination
+                # So its offset should be 0
+                audios_idx[key]["src_to_offset"][(src_chunk_idx, src_file_idx)] = 0
+                chunk_idx, file_idx = update_chunk_file_indices(chunk_idx, file_idx, chunk_size)
+                dst_path = dst_meta.root / DEFAULT_AUDIO_PATH.format(
+                    audio_key=key,
+                    chunk_index=chunk_idx,
+                    file_index=file_idx,
+                )
+                dst_path.parent.mkdir(parents=True, exist_ok=True)
+                shutil.copy(str(src_path), str(dst_path))
+                # Reset offset for next file
+                current_offset = src_duration
+            else:
+                # Append to existing video file - use current accumulated offset
+                audios_idx[key]["src_to_offset"][(src_chunk_idx, src_file_idx)] = current_offset
+                concatenate_media_files(
+                    [dst_path, src_path],
+                    dst_path,
+                )
+                current_offset += src_duration
+
+            audios_idx[key]["episode_duration"] += src_duration
+
+        audios_idx[key]["chunk"] = chunk_idx
+        audios_idx[key]["file"] = file_idx
+
+    return audios_idx
+
+
 def aggregate_data(src_meta, dst_meta, data_idx, data_files_size_in_mb, chunk_size):
     """Aggregates data chunks from a source dataset into the destination dataset.
 
@@ -497,7 +639,7 @@ def aggregate_data(src_meta, dst_meta, data_idx, data_files_size_in_mb, chunk_si
     return data_idx
 
 
-def aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx):
+def aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx, audios_idx):
     """Aggregates metadata from a source dataset into the destination dataset.
 
     Reads source metadata files, updates all indices and timestamps,
@@ -509,6 +651,7 @@ def aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx):
         meta_idx: Dictionary tracking metadata chunk and file indices.
         data_idx: Dictionary tracking data chunk and file indices.
         videos_idx: Dictionary tracking video indices and timestamps.
+        audios_idx: Dictionary tracking audio indices and timestamps.
 
     Returns:
         dict: Updated meta_idx with current chunk and file indices.
@@ -532,6 +675,7 @@ def aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx):
             meta_idx,
             data_idx,
             videos_idx,
+            audios_idx,
         )
 
         meta_idx, _ = append_or_create_parquet_file(
@@ -548,7 +692,8 @@ def aggregate_metadata(src_meta, dst_meta, meta_idx, data_idx, videos_idx):
     # Increment latest_duration by the total duration added from this source dataset
     for k in videos_idx:
         videos_idx[k]["latest_duration"] += videos_idx[k]["episode_duration"]
-
+    for k in audios_idx:
+        audios_idx[k]["latest_duration"] += audios_idx[k]["episode_duration"]
     return meta_idx
 
 

src/lerobot/datasets/audio_utils.py

@@ -0,0 +1,275 @@
+#!/usr/bin/env python
+
+# Copyright 2025 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 logging
+from pathlib import Path
+
+import av
+import torch
+import torchaudio
+import torchcodec
+from numpy import ceil
+
+CHANNELS_LAYOUTS_MAPPING = {
+    1: "mono",
+    2: "stereo",
+    3: "2.1",
+    4: "3.1",
+    5: "4.1",
+    6: "5.1",
+    7: "6.1",
+    8: "7.1",
+    16: "hexadecagonal",
+    24: "22.2",
+}
+
+
+def decode_audio(
+    audio_path: Path | str,
+    timestamps: list[float],
+    duration: float,
+    start_time_s: float | None = 0.0,
+    backend: str | None = "torchcodec",
+) -> torch.Tensor:
+    """
+    Decodes audio using the specified backend.
+    Args:
+        audio_path (Path): Path to the audio file.
+        timestamps (list[float]): List of (starting) timestamps to extract audio chunks.
+        duration (float): Duration of the audio chunks in seconds.
+        backend (str, optional): Backend to use for decoding. Defaults to "torchcodec".
+
+    Returns:
+        torch.Tensor: Decoded audio chunks.
+
+    Currently supports torchaudio.
+    """
+    if backend == "torchcodec":
+        return decode_audio_torchcodec(audio_path, timestamps, duration, start_time_s)
+    elif backend == "torchaudio":
+        return decode_audio_torchaudio(audio_path, timestamps, duration, start_time_s)
+    else:
+        raise ValueError(f"Unsupported video backend: {backend}")
+
+
+def decode_audio_torchcodec(
+    audio_path: Path | str,
+    timestamps: list[float],
+    duration: float,
+    start_time_s: float | None = 0.0,
+    log_loaded_timestamps: bool = False,
+) -> torch.Tensor:
+    # TODO(CarolinePascal) : add channels selection
+    audio_decoder = torchcodec.decoders.AudioDecoder(audio_path)
+    audio_sample_rate = audio_decoder.metadata.sample_rate
+    audio_channels = audio_decoder.metadata.num_channels
+    # TODO(CarolinePascal) : assert ts < total record duration
+
+    audio_chunks = []
+    timestamps = [
+        timestamp + start_time_s for timestamp in timestamps
+    ]  # Add an offset of start_time_s to each timestamp
+    for ts in timestamps:
+        current_audio_chunk = audio_decoder.get_samples_played_in_range(
+            start_seconds=max(0.0, ts - duration), stop_seconds=ts
+        )
+
+        current_audio_chunk_data = current_audio_chunk.data
+
+        # Case where the requested audio chunk starts before the beginning of the audio stream
+        if ts - duration < 0:
+            # No useful audio sample has been recorded
+            if ts < 1 / audio_sample_rate:
+                # TODO(CarolinePascal) : add low level white noise instead of zeros ?
+                current_audio_chunk_data = torch.zeros(
+                    (audio_channels, int(ceil(duration * audio_sample_rate)))
+                )
+            # At least one useful audio sample has been recorded
+            else:
+                # Pad the beginning of the audio chunk with zeros
+                # TODO(CarolinePascal) : add low level white noise instead of zeros ?
+                current_audio_chunk_data = torch.nn.functional.pad(
+                    current_audio_chunk_data,
+                    (int(ceil((duration - ts) * audio_sample_rate)), 0, 0, 0),  # left, right, top, bottom
+                )
+
+        if log_loaded_timestamps:
+            logging.info(
+                f"audio chunk loaded at timestamp={current_audio_chunk.pts_seconds:.4f} with duration={current_audio_chunk.duration_seconds:.4f}"
+            )
+
+        audio_chunks.append(current_audio_chunk_data)
+
+    audio_chunks = torch.stack(audio_chunks)
+
+    assert len(timestamps) == len(audio_chunks)
+    return audio_chunks
+
+
+def decode_audio_torchaudio(
+    audio_path: Path | str,
+    timestamps: list[float],
+    duration: float,
+    start_time_s: float | None = 0.0,
+    log_loaded_timestamps: bool = False,
+) -> torch.Tensor:
+    # TODO(CarolinePascal) : add channels selection
+    audio_path = str(audio_path)
+
+    reader = torchaudio.io.StreamReader(src=audio_path)
+    audio_sample_rate = reader.get_src_stream_info(reader.default_audio_stream).sample_rate
+    audio_channels = reader.get_src_stream_info(reader.default_audio_stream).num_channels
+    # TODO(CarolinePascal) : assert ts < total record duration
+
+    # TODO(CarolinePascal) : sort timestamps ?
+
+    reader.add_basic_audio_stream(
+        frames_per_chunk=int(ceil(duration * audio_sample_rate)),  # Too much is better than not enough
+        buffer_chunk_size=-1,  # No dropping frames
+        format="fltp",  # Format as float32
+    )
+
+    audio_chunks = []
+    timestamps = [
+        timestamp + start_time_s for timestamp in timestamps
+    ]  # Add an offset of start_time_s to each timestamp
+    for ts in timestamps:
+        reader.seek(max(0.0, ts - duration))  # Default to closest audio sample. Needs to be non-negative !
+        status = reader.fill_buffer()
+        if status != 0:
+            # Should not happen, but just in case
+            logging.warning("Audio stream reached end of recording before decoding desired timestamps.")
+
+        current_audio_chunk = reader.pop_chunks()[0]
+        current_audio_chunk_data = current_audio_chunk.t()  # Channel first format
+
+        # Case where the requested audio chunk starts before the beginning of the audio stream
+        if ts - duration < 0:
+            # No useful audio sample has been recorded
+            if ts < 1 / audio_sample_rate:
+                current_audio_chunk_data = torch.zeros(
+                    (audio_channels, int(ceil(duration * audio_sample_rate)))
+                )
+            # At least one useful audio sample has been recorded
+            else:
+                # Remove the superfluous last samples of the audio chunk
+                current_audio_chunk_data = current_audio_chunk_data[:, : int(ceil(ts * audio_sample_rate))]
+                # Pad the beginning of the audio chunk with zeros
+                # TODO(CarolinePascal) : add low level white noise instead of zeros ?
+                current_audio_chunk_data = torch.nn.functional.pad(
+                    current_audio_chunk_data,
+                    (int(ceil((duration - ts) * audio_sample_rate)), 0, 0, 0),  # left, right, top, bottom
+                )
+
+        if log_loaded_timestamps:
+            logging.info(
+                f"audio chunk loaded at starting timestamp={current_audio_chunk['pts']:.4f} with duration={len(current_audio_chunk) / audio_sample_rate:.4f}"
+            )
+
+        audio_chunks.append(current_audio_chunk_data)
+
+    audio_chunks = torch.stack(audio_chunks)
+
+    assert len(timestamps) == len(audio_chunks)
+    return audio_chunks
+
+
+def encode_audio(
+    input_path: Path | str,
+    output_path: Path | str,
+    codec: str = "aac",  # TODO(CarolinePascal) : investigate Fraunhofer FDK AAC (libfdk_aac) codec and and constant (file size control) /variable (quality control) bitrate options
+    bit_rate: int | None = None,
+    sample_rate: int | None = None,
+    log_level: int | None = av.logging.ERROR,
+    overwrite: bool = False,
+) -> None:
+    """Encodes an audio file using ffmpeg."""
+    output_path = Path(output_path)
+    output_path.parent.mkdir(parents=True, exist_ok=overwrite)
+
+    # Set logging level
+    if log_level is not None:
+        # "While less efficient, it is generally preferable to modify logging with Python’s logging"
+        logging.getLogger("libav").setLevel(log_level)
+
+    # Open input file
+    with av.open(str(input_path), "r") as input:
+        input_stream = input.streams.audio[0]  # Assuming the first stream is the audio stream to be encoded
+
+        # Define sub-sampling options
+        if sample_rate is None:
+            sample_rate = input_stream.rate
+
+        # Create and open output file (overwrite by default)
+        with av.open(str(output_path), "w") as output:
+            output_stream = output.add_stream(
+                codec, rate=sample_rate, layout=CHANNELS_LAYOUTS_MAPPING[input_stream.channels]
+            )
+
+            if bit_rate is not None:
+                output_stream.bit_rate = bit_rate
+
+            # Loop through input WAV packets and encode them
+            for input_frame in input.decode(
+                input_stream
+            ):  # This step handles both demuxing and decoding under the hood
+                packet = output_stream.encode(input_frame)
+                if packet:
+                    output.mux(packet)
+
+            # Flush the encoder
+            packet = output_stream.encode()
+            if packet:
+                output.mux(packet)
+
+    # Reset logging level
+    if log_level is not None:
+        av.logging.restore_default_callback()
+
+    if not output_path.exists():
+        raise OSError(f"Audio encoding did not work. File not found: {output_path}.")
+
+
+def get_audio_info(video_path: Path | str) -> dict:
+    # Set logging level
+    logging.getLogger("libav").setLevel(av.logging.ERROR)
+
+    # Getting audio stream information
+    audio_info = {}
+    with av.open(str(video_path), "r") as audio_file:
+        try:
+            audio_stream = audio_file.streams.audio[0]
+        except IndexError:
+            # Reset logging level
+            av.logging.restore_default_callback()
+            return {"has_audio": False}
+
+        audio_info["audio.channels"] = audio_stream.channels
+        audio_info["audio.codec"] = audio_stream.codec.canonical_name
+        # In an ideal loseless case : bit depth x sample rate x channels = bit rate.
+        # In an actual compressed case, the bit rate is set according to the compression level : the lower the bit rate, the more compression is applied.
+        audio_info["audio.bit_rate"] = audio_stream.bit_rate
+        audio_info["audio.sample_rate"] = audio_stream.sample_rate  # Number of samples per second
+        # In an ideal loseless case : fixed number of bits per sample.
+        # In an actual compressed case : variable number of bits per sample (often reduced to match a given depth rate).
+        audio_info["audio.bit_depth"] = audio_stream.format.bits
+        audio_info["audio.channel_layout"] = audio_stream.layout.name
+        audio_info["has_audio"] = True
+
+    # Reset logging level
+    av.logging.restore_default_callback()
+
+    return audio_info

src/lerobot/datasets/backward_compatibility.py

@@ -1,56 +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 packaging.version
-
-V30_MESSAGE = """
-The dataset you requested ({repo_id}) is in {version} format.
-
-We introduced a new format since v3.0 which is not backward compatible with v2.1.
-Please, update your dataset to the new format using this command:
-```
-python -m lerobot.datasets.v30.convert_dataset_v21_to_v30 --repo-id={repo_id}
-```
-
-If you already have a converted version uploaded to the hub, then this error might be because of
-an older version in your local cache. Consider deleting the cached version and retrying.
-
-If you encounter a problem, contact LeRobot maintainers on [Discord](https://discord.com/invite/s3KuuzsPFb)
-or open an [issue on GitHub](https://github.com/huggingface/lerobot/issues/new/choose).
-"""
-
-FUTURE_MESSAGE = """
-The dataset you requested ({repo_id}) is only available in {version} format.
-As we cannot ensure forward compatibility with it, please update your current version of lerobot.
-"""
-
-
-class CompatibilityError(Exception): ...
-
-
-class BackwardCompatibilityError(CompatibilityError):
-    def __init__(self, repo_id: str, version: packaging.version.Version):
-        if version.major == 2 and version.minor == 1:
-            message = V30_MESSAGE.format(repo_id=repo_id, version=version)
-        else:
-            raise NotImplementedError(
-                "Contact the maintainer on [Discord](https://discord.com/invite/s3KuuzsPFb)."
-            )
-        super().__init__(message)
-
-
-class ForwardCompatibilityError(CompatibilityError):
-    def __init__(self, repo_id: str, version: packaging.version.Version):
-        message = FUTURE_MESSAGE.format(repo_id=repo_id, version=version)
-        super().__init__(message)

src/lerobot/datasets/card_template.md

@@ -7,6 +7,13 @@
 
 This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
 
+{% if repo_id is defined and repo_id %}
+<a class="flex" href="https://huggingface.co/spaces/lerobot/visualize_dataset?path={{ repo_id }}">
+<img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl.svg"/>
+<img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl-dark.svg"/>
+</a>
+{% endif %}
+
 ## Dataset Description
 
 {{ dataset_description | default("", true) }}

src/lerobot/datasets/compute_stats.py

@@ -13,9 +13,13 @@
 # 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 __future__ import annotations
+
+import logging
+
 import numpy as np
 
-from lerobot.datasets.utils import load_image_as_numpy
+from lerobot.datasets.io_utils import load_audio_from_path, load_image_as_numpy
 
 DEFAULT_QUANTILES = [0.01, 0.10, 0.50, 0.90, 0.99]
 
@@ -245,6 +249,20 @@ def sample_images(image_paths: list[str]) -> np.ndarray:
     return images
 
 
+def sample_audio_from_path(audio_path: str) -> np.ndarray:
+    """Samples audio data from an audio recording stored in a WAV file."""
+    data = load_audio_from_path(audio_path)
+    sampled_indices = sample_indices(len(data))
+
+    return data[sampled_indices]
+
+
+def sample_audio_from_data(data: np.ndarray) -> np.ndarray:
+    """Samples audio data from an audio recording stored in a numpy array."""
+    sampled_indices = sample_indices(len(data))
+    return data[sampled_indices]
+
+
 def _reshape_stats_by_axis(
     stats: dict[str, np.ndarray],
     axis: int | tuple[int, ...] | None,
@@ -512,6 +530,13 @@ def compute_episode_stats(
             ep_ft_array = sample_images(data)
             axes_to_reduce = (0, 2, 3)
             keepdims = True
+        elif features[key]["dtype"] == "audio":
+            try:
+                ep_ft_array = sample_audio_from_path(data[0])
+            except TypeError:  # Should only be triggered for LeKiwi robot, for which audio is stored chunk by chunk in a visual frame-like manner
+                ep_ft_array = sample_audio_from_data(data)
+            axes_to_reduce = 0
+            keepdims = True
         else:
             ep_ft_array = data
             axes_to_reduce = 0
@@ -624,3 +649,141 @@ def aggregate_stats(stats_list: list[dict[str, dict]]) -> dict[str, dict[str, np
         aggregated_stats[key] = aggregate_feature_stats(stats_with_key)
 
     return aggregated_stats
+
+
+def _get_valid_chunk_starts(episode_indices: np.ndarray, chunk_size: int) -> np.ndarray:
+    """Return all start indices where a chunk of ``chunk_size`` stays within one episode."""
+    total = len(episode_indices)
+    if total < chunk_size:
+        return np.array([], dtype=np.int64)
+    max_start = total - chunk_size
+    starts = np.arange(max_start + 1)
+    valid = episode_indices[starts] == episode_indices[starts + chunk_size - 1]
+    return starts[valid]
+
+
+def _compute_relative_chunk_batch(
+    start_indices: np.ndarray,
+    all_actions: np.ndarray,
+    all_states: np.ndarray,
+    chunk_size: int,
+    relative_mask: np.ndarray,
+) -> np.ndarray:
+    """Vectorised relative-action computation for a batch of start indices.
+
+    Returns an ``(N * chunk_size, action_dim)`` float32 array.
+    """
+    if len(start_indices) == 0:
+        return np.empty((0, all_actions.shape[1]), dtype=np.float32)
+    offsets = np.arange(chunk_size)
+    frame_idx = start_indices[:, None] + offsets[None, :]
+    chunks = all_actions[frame_idx].copy()
+    states = all_states[start_indices]
+    mask_dim = len(relative_mask)
+    chunks[:, :, :mask_dim] -= states[:, None, :mask_dim] * relative_mask[None, None, :]
+    return chunks.reshape(-1, all_actions.shape[1])
+
+
+def compute_relative_action_stats(
+    hf_dataset,
+    features: dict,
+    chunk_size: int,
+    exclude_joints: list[str] | None = None,
+    num_workers: int = 0,
+) -> dict[str, np.ndarray]:
+    """Compute normalization statistics for relative actions over the full dataset.
+
+    Iterates *all* valid action chunks (within single episodes), converts them to
+    relative actions (action − current_state), and computes per-dimension
+    statistics suitable for normalization.
+
+    Args:
+        hf_dataset: The underlying HuggingFace dataset with "action",
+            "observation.state", and "episode_index" columns.
+        features: Dataset feature metadata (must contain "action" with "shape"
+            and optionally "names").
+        chunk_size: Number of consecutive frames per action chunk.
+        exclude_joints: Joint names whose dimensions should remain absolute
+            (not converted to relative actions).
+        num_workers: Number of parallel threads for computation. Values ≤1
+            mean single-threaded. Numpy releases the GIL so threads give
+            real parallelism here.
+
+    Returns:
+        Statistics dict with keys "mean", "std", "min", "max", "q01", …, "q99".
+
+    Raises:
+        ValueError: If the dataset has fewer frames than ``chunk_size``.
+        RuntimeError: If no valid (single-episode) chunks are found.
+    """
+    from lerobot.processor.relative_action_processor import RelativeActionsProcessorStep
+
+    if exclude_joints is None:
+        exclude_joints = []
+
+    action_dim = features[ACTION]["shape"][0]
+    action_names = features.get(ACTION, {}).get("names")
+    mask_step = RelativeActionsProcessorStep(
+        enabled=True,
+        exclude_joints=exclude_joints,
+        action_names=action_names,
+    )
+    relative_mask = np.array(mask_step._build_mask(action_dim), dtype=np.float32)
+
+    logging.info("Loading action/state data for relative action stats...")
+    all_actions = np.array(hf_dataset[ACTION], dtype=np.float32)
+    all_states = np.array(hf_dataset[OBS_STATE], dtype=np.float32)
+    episode_indices = np.array(hf_dataset["episode_index"])
+
+    valid_starts = _get_valid_chunk_starts(episode_indices, chunk_size)
+    if len(valid_starts) == 0:
+        raise RuntimeError(
+            f"No valid chunks found (total_frames={len(episode_indices)}, chunk_size={chunk_size})"
+        )
+
+    effective_workers = max(num_workers, 1)
+    logging.info(
+        f"Computing relative action stats from {len(valid_starts)} chunks "
+        f"(chunk_size={chunk_size}, workers={effective_workers})"
+    )
+
+    batch_size = 50_000
+    batches = [valid_starts[i : i + batch_size] for i in range(0, len(valid_starts), batch_size)]
+
+    running_stats = RunningQuantileStats()
+
+    if num_workers > 1:
+        from concurrent.futures import ThreadPoolExecutor, as_completed
+
+        with ThreadPoolExecutor(max_workers=num_workers) as pool:
+            futures = [
+                pool.submit(
+                    _compute_relative_chunk_batch,
+                    batch,
+                    all_actions,
+                    all_states,
+                    chunk_size,
+                    relative_mask,
+                )
+                for batch in batches
+            ]
+            for future in as_completed(futures):
+                running_stats.update(future.result())
+    else:
+        for batch in batches:
+            running_stats.update(
+                _compute_relative_chunk_batch(batch, all_actions, all_states, chunk_size, relative_mask)
+            )
+
+    stats = running_stats.get_statistics()
+
+    excluded_dims = int(len(relative_mask) - relative_mask.sum())
+    total_frames = len(valid_starts) * chunk_size
+    logging.info(
+        f"Relative action stats ({len(valid_starts)} chunks, {total_frames} frames): "
+        f"relative_dims={int(relative_mask.sum())}/{len(relative_mask)} (excluded={excluded_dims}), "
+        f"mean={np.abs(stats['mean']).mean():.4f}, std={stats['std'].mean():.4f}, "
+        f"q01={stats['q01'].mean():.4f}, q99={stats['q99'].mean():.4f}"
+    )
+
+    return stats