Merge branch 'main' into feat/sonic_encoder_decoder

docs/source/annotation_pipeline.mdx

@@ -89,8 +89,8 @@ subtask.
 
 The resulting spans are then stitched into a gap-free, full-episode
 cover, so **every frame has exactly one active subtask**. See
-[`run_hf_job.py`](https://github.com/huggingface/lerobot/blob/main/examples/annotations/run_hf_job.py)
-for the production settings (single camera, timestamped contact sheets,
+[Running on Hugging Face Jobs](#running-on-hugging-face-jobs) for the
+production settings (single camera, timestamped contact sheets,
 auto-windowed subtask generation).
 
 ### Tools
@@ -110,28 +110,67 @@ not-yet-implemented.
 
 ## Running on Hugging Face Jobs
 
-Annotation runs on [Hugging Face Jobs](https://huggingface.co/docs/hub/en/jobs).
-The repo ships a launcher script you copy and tweak for your dataset:
+Annotating a real dataset needs a GPU big enough to serve the VLM, so
+`lerobot-annotate` can dispatch itself to
+[Hugging Face Jobs](https://huggingface.co/docs/hub/en/jobs) — same as
+`lerobot-train`. Add `--job.target=<flavor>` to the exact command you'd
+run locally and it runs on that hardware instead:
 
 ```bash
-HF_TOKEN=hf_... uv run python examples/annotations/run_hf_job.py
+hf auth login   # once
+
+uv run lerobot-annotate \
+    --repo_id=user/my_dataset \
+    --new_repo_id=user/my_dataset_annotated \
+    --push_to_hub=true \
+    --vlm.model_id=Qwen/Qwen3.6-27B \
+    --vlm.num_gpus=1 \
+    --vlm.serve_command="vllm serve Qwen/Qwen3.6-27B --tensor-parallel-size 1 \
+        --max-model-len 32768 --gpu-memory-utilization 0.8 \
+        --uvicorn-log-level warning --port {port}" \
+    --vlm.serve_ready_timeout_s=1800 \
+    --vlm.chat_template_kwargs='{"enable_thinking": false}' \
+    --job.target=h200
 ```
 
-[`run_hf_job.py`](https://github.com/huggingface/lerobot/blob/main/examples/annotations/run_hf_job.py)
-starts a single-GPU `h200` job (bump it to `h200x4` for big datasets)
-that:
+That submits a single-GPU `h200` job that:
 
-1. installs `lerobot` (from `main`) plus the annotation extras,
-2. boots one vLLM server per GPU (using the `vllm/vllm-openai` image) and
-   drives it over the OpenAI-compatible API,
-3. runs the `plan` / `interjections` / `vqa` modules across the dataset
-   with `lerobot-annotate`,
+1. starts from the `vllm/vllm-openai` image and installs `lerobot` on top,
+2. boots one vLLM server per GPU and drives it over the OpenAI-compatible API,
+3. runs the `plan` / `interjections` / `vqa` modules across the dataset,
 4. with `--push_to_hub=true`, uploads the result to `--new_repo_id` (or
    back to `--repo_id` in place if you leave that unset).
 
-To use a different dataset, model, or hub repo, edit the `CMD` block in
-the script. Every flag there maps directly to a `lerobot-annotate` flag
-(run `lerobot-annotate --help` for the full list).
+The command streams the job's logs; `Ctrl-C` detaches without cancelling
+it. List the available flavors and their pricing with `hf jobs hardware`.
+
+<Tip warning={true}>
+
+Qwen3.6 ships with thinking enabled, which eats the token budget the
+annotator needs for its JSON answer — `--vlm.chat_template_kwargs='{"enable_thinking": false}'`
+turns it off. Without `--push_to_hub=true` the annotated dataset is
+discarded when the pod exits.
+
+</Tip>
+
+### Job options
+
+| Flag                | Default                   | What it does                                                                    |
+| ------------------- | ------------------------- | ------------------------------------------------------------------------------- |
+| `--job.target`      | `local`                   | HF Jobs flavor to run on (e.g. `h200`, `h200x4`). Omitted/`local` runs here.    |
+| `--job.image`       | `vllm/vllm-openai:latest` | Runtime image for the pod.                                                      |
+| `--job.timeout`     | `2h`                      | Wall-clock cap. Raise it for large datasets.                                    |
+| `--job.detach`      | `false`                   | Submit and exit instead of streaming logs.                                      |
+| `--job.lerobot_ref` | `main`                    | Git ref of lerobot installed on the pod — point it at a branch to test changes. |
+| `--job.tags`        | `[]`                      | Extra tags on the job and on any dataset it pushes (`lerobot` is always added). |
+
+For a bigger dataset, scale to `h200x4` and raise
+`--vlm.parallel_servers` / `--vlm.num_gpus` to match, and give the job
+more headroom with e.g. `--job.timeout=8h`.
+
+Remote runs need `--repo_id` (the pod pulls the dataset from the Hub;
+`--root` names a directory only your machine has). A dataset that exists
+only in your local cache is pushed to a **private** repo first.
 
 ## Key options
 

docs/source/backwardcomp.mdx

@@ -58,7 +58,7 @@ final_action = postprocessor(action)
 
 ## Hardware API redesign
 
-PR [#777](https://github.com/huggingface/lerobot/pull/777) improves the LeRobot calibration but is **not backward-compatible**. Below is a overview of what changed and how you can continue to work with datasets created before this pull request.
+PR [#777](https://github.com/huggingface/lerobot/pull/777) improves the LeRobot calibration but is **not backward-compatible**. Below is an overview of what changed and how you can continue to work with datasets created before this pull request.
 
 ### What changed?
 
@@ -129,8 +129,8 @@ python examples/backward_compatibility/replay.py \
 
 Policies output actions in the same format as the datasets (`torch.Tensors`). Therefore, the same transformations should be applied.
 
-To find these transformations, we recommend to first try and and replay an episode of the dataset your policy was trained on using the section above.
-Then, add these same transformations on your inference script (shown here in the `record.py` script):
+To find these transformations, we recommend first replaying an episode of the dataset your policy was trained on using the section above.
+Then, add these same transformations to your inference script (shown here in the `record.py` script):
 
 ```diff
 action_values = predict_action(

docs/source/bring_your_own_policies.mdx

@@ -165,6 +165,8 @@ Batches are flat dictionaries keyed by the constants in [`lerobot.utils.constant
 
 LeRobot uses `PolicyProcessorPipeline`s to normalize inputs and de-normalize outputs around your policy. 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).
 
+Pay close attention here: processors are the most common reproducibility pain point. A mismatch in normalization mode (`IDENTITY` vs `MEAN_STD` vs `MIN_MAX` vs `QUANTILES`/`QUANTILE10`) or in which features get normalized will train and eval without erroring, yet silently wreck results. Make sure the modes match how the checkpoint was trained, that the required stats exist (e.g. `QUANTILES` needs `q01`/`q99`), and that the pre- and post-processors stay consistent.
+
 ```python
 # processor_my_policy.py
 from typing import Any
@@ -304,7 +306,9 @@ Mirror an existing policy that's structurally similar to yours; the diff is smal
 
 ### Heavy / optional dependencies
 
-Most policies need a heavy backbone (transformers, diffusers, a specific VLM SDK). The convention is **two-step gating**: a `TYPE_CHECKING`-guarded import at module top, and a `require_package` runtime check in the constructor. [`modeling_diffusion.py`](https://github.com/huggingface/lerobot/blob/main/src/lerobot/policies/diffusion/modeling_diffusion.py) is the canonical reference:
+Most policies need a heavy backbone (transformers, diffusers, a specific VLM SDK). Wherever one exists, prefer loading it e.g from `transformers` or `diffusers` rather than re-implementing the architecture in-tree.
+
+The convention is **two-step gating**: a `TYPE_CHECKING`-guarded import at module top, and a `require_package` runtime check in the constructor. [`modeling_diffusion.py`](https://github.com/huggingface/lerobot/blob/main/src/lerobot/policies/diffusion/modeling_diffusion.py) is the canonical reference:
 
 ```python
 from typing import TYPE_CHECKING
@@ -374,6 +378,7 @@ The general expectations are in [`CONTRIBUTING.md`](https://github.com/huggingfa
 - [ ] Optional deps live behind a `[project.optional-dependencies]` extra and the `TYPE_CHECKING + require_package` guard.
 - [ ] `tests/policies/` updated; backward-compat artifact committed & policy-specific tests.
 - [ ] `src/lerobot/policies/<name>/README.md` symlinked into `docs/source/policy_<name>_README.md`; user-facing `docs/source/<name>.mdx` written and added to `_toctree.yml`.
+- [ ] `lerobot-train --policy.type my_policy ...` runs end-to-end for at least a few steps + save a checkpoint that can be loaded and run by `lerobot-eval` or `lerobot-rollout`.
 - [ ] `templates/lerobot_modelcard_template.md` has a description entry and a `policy_docs` link for your policy.
 - [ ] The models table in the root `README.md` lists your policy in the right category, linking to your doc page.
 - [ ] At least one reproducible benchmark eval in the policy MDX with a published checkpoint (sim benchmark, or real-robot dataset + checkpoint).

docs/source/cameras.mdx

@@ -136,6 +136,10 @@ config = RealSenseCameraConfig(
     height=480,
     color_mode=ColorMode.RGB,
     use_depth=True,
+    # Optional fixed color controls. Omit them to leave the current sensor settings unchanged.
+    exposure=120,
+    gain=64,
+    white_balance=4600,
     rotation=Cv2Rotation.NO_ROTATION
 )
 
@@ -154,6 +158,15 @@ finally:
 ```
 <!-- prettier-ignore-end -->
 
+Manual color controls disable the corresponding automatic exposure or white-balance mode. Their
+supported ranges vary by camera model; an invalid value raises an error at connection time that
+includes the range reported by the sensor. Requesting an unsupported control also raises an error.
+Omitted controls leave the sensor's existing automatic or manual setting unchanged. These options
+require `use_rgb=True`.
+
+On the RealSense D405, the color stream is provided by the Stereo Module, so changing manual
+exposure or gain also affects the depth stream.
+
 </hfoption>
 </hfoptions>
 

docs/source/env_processor.mdx

@@ -88,20 +88,6 @@ policy_preprocessor = NormalizerProcessorStep(stats=dataset_stats)
 
 The same policy can work with different environment processors, and the same environment processor can work with different policies:
 
-````python
-# Use SmolVLA policy with LIBERO environment
-# Use SmolVLA policy with LIBERO environment
-libero_preprocessor, libero_postprocessor = make_env_pre_post_processors(
-    env_cfg=libero_cfg,
-    policy_cfg=smolvla_cfg,
-)
-smolvla_preprocessor, smolvla_postprocessor = make_pre_post_processors(smolvla_cfg)
-# Or use ACT policy with the same LIBERO environment
-libero_preprocessor, libero_postprocessor = make_env_pre_post_processors(
-    env_cfg=libero_cfg,
-    policy_cfg=act_cfg,
-)
-act_preprocessor, act_postprocessor = make_pre_post_processors(act_cfg)
 ```python
 # Use SmolVLA policy with LIBERO environment
 libero_preprocessor, libero_postprocessor = make_env_pre_post_processors(
@@ -116,6 +102,7 @@ libero_preprocessor, libero_postprocessor = make_env_pre_post_processors(
     policy_cfg=act_cfg,
 )
 act_preprocessor, act_postprocessor = make_pre_post_processors(act_cfg)
+```
 
 ### 3. **Easier Experimentation**
 
@@ -145,7 +132,7 @@ class LiberoVelocityProcessorStep(ObservationProcessorStep):
         state = torch.cat([eef_pos, eef_axisangle, eef_vel,
                           gripper_pos, gripper_vel], dim=-1)  # 14D
         return state
-````
+```
 
 ### 4. **Cleaner Environment Code**
 

docs/source/feetech.mdx

@@ -40,10 +40,10 @@ This tutorial guides you through updating the firmware of Feetech motors using t
 For each motor you want to update:
 
 1. **Select the motor** from the list by clicking on it
-2. **Click on Upgrade tab**:
-3. **Click on Online button**:
-   - If an potential firmware update is found, it will be displayed in the box
-4. **Click on Upgrade button**:
+2. **Click the Upgrade tab**:
+3. **Click the Online button**:
+   - If a potential firmware update is found, it will be displayed in the box
+4. **Click the Upgrade button**:
    - The update progress will be displayed
 
 ## Step 6: Verify Update

docs/source/hope_jr.mdx

@@ -211,7 +211,7 @@ Record, Replay and Train with Hope-JR is still experimental.
 
 ### Record
 
-This step records the dataset, which can be seen as an example [here](https://huggingface.co/datasets/nepyope/hand_record_test_with_video_data/settings).
+This step records the dataset, which can be seen as an example [here](https://huggingface.co/datasets/nepyope/hand_record_test_with_video_data).
 
 ```bash
 lerobot-record \

docs/source/integrate_hardware.mdx

@@ -18,7 +18,7 @@ If you're using Feetech or Dynamixel motors, LeRobot provides built-in bus inter
 - [`DynamixelMotorsBus`](https://github.com/huggingface/lerobot/blob/main/src/lerobot/motors/dynamixel/dynamixel.py) – for controlling Dynamixel servos
 
 Please refer to the [`MotorsBus`](https://github.com/huggingface/lerobot/blob/main/src/lerobot/motors/motors_bus.py) abstract class to learn about its API.
-For a good example of how it can be used, you can have a look at our own [SO101 follower implementation](https://github.com/huggingface/lerobot/blob/main/src/lerobot/robots/so_follower/so101_follower/so101_follower.py)
+For a good example of how it can be used, you can have a look at our own [SO101 follower implementation](https://github.com/huggingface/lerobot/blob/main/src/lerobot/robots/so_follower/so_follower.py)
 
 Use these if compatible. Otherwise, you'll need to find or write a Python interface (not covered in this tutorial):
 

docs/source/lekiwi.mdx

@@ -51,7 +51,7 @@ In addition to these instructions, you need to install the Feetech SDK & ZeroMQ
 pip install -e ".[lekiwi]"
 ```
 
-Great :hugs:! You are now done installing LeRobot, and we can begin assembling the SO100/SO101 arms and the mobile base :robot:.
+Great 🤗! You are now done installing LeRobot, and we can begin assembling the SO100/SO101 arms and the mobile base 🤖.
 Every time you now want to use LeRobot, you can go to the `~/lerobot` folder where we installed LeRobot and run one of the commands.
 
 # Step-by-Step Assembly Instructions

docs/source/omx.mdx

@@ -1,3 +1,11 @@
+# OMX
+
+<img
+  src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/omx_mainimage.png"
+  alt="OMX"
+  width=600
+/>
+
 ## Order and Assemble the parts
 
 First, assemble the OMX hardware following the official assembly guide.

docs/source/pi0fast.mdx

@@ -174,7 +174,7 @@ The model takes images, text instructions, and robot state as input, and outputs
 
 ## Reproducing π₀Fast results
 
-We reproduce the results of π₀Fast on the LIBERO benchmark using the LeRobot implementation. We take the LeRobot PiFast base model [lerobot/pi0fast-base](https://huggingface.co/lerobot/pi0fast-base) and finetune for an additional 40kk steps in bfloat16, with batch size of 256 on 8 H100 GPUs using the [HuggingFace LIBERO dataset](https://huggingface.co/datasets/HuggingFaceVLA/libero).
+We reproduce the results of π₀Fast on the LIBERO benchmark using the LeRobot implementation. We take the LeRobot PiFast base model [lerobot/pi0fast-base](https://huggingface.co/lerobot/pi0fast-base) and finetune for an additional 40k steps in bfloat16, with batch size of 256 on 8 H100 GPUs using the [HuggingFace LIBERO dataset](https://huggingface.co/datasets/HuggingFaceVLA/libero).
 
 The finetuned model can be found here:
 

docs/source/processors_robots_teleop.mdx

@@ -22,7 +22,7 @@ With processors, you choose the learning features you want to use for your polic
 ## Three pipelines
 
 We often compose three pipelines. Depending on your setup, some can be empty if action and observation spaces already match.
-Each of these pipelines handle different conversions between different action and observation spaces. Below is a quick explanation of each pipeline.
+Each of these pipelines handles different conversions between different action and observation spaces. Below is a quick explanation of each pipeline.
 
 1. Pipeline 1: Teleop action space → dataset action space (phone pose → EE targets)
 2. Pipeline 2: Dataset action space → robot command space (EE targets → joints)
@@ -74,15 +74,15 @@ In the phone to SO-100 follower examples we use the following adapters:
 - `robot_action_to_transition`: transforms the teleop action dict to a pipeline transition.
 - `transition_to_robot_action`: transforms the pipeline transition to a robot action dict.
 - `observation_to_transition`: transforms the robot observation dict to a pipeline transition.
-- `transition_to_observation`: transforms the pipeline transition to a observation dict.
+- `transition_to_observation`: transforms the pipeline transition to an observation dict.
 
-Checkout [src/lerobot/processor/converters.py](https://github.com/huggingface/lerobot/blob/main/src/lerobot/processor/converters.py) for more details.
+Check out [src/lerobot/processor/converters.py](https://github.com/huggingface/lerobot/blob/main/src/lerobot/processor/converters.py) for more details.
 
 ## Dataset feature contracts
 
 Dataset features are determined by the keys saved in the dataset. Each step can declare what features it modifies in a contract called `transform_features(...)`. Once you build a processor, the processor can then aggregate all of these features with `aggregate_pipeline_dataset_features()` and merge multiple feature dicts with `combine_feature_dicts(...)`.
 
-Below is and example of how we declare features with the `transform_features` method in the phone to SO-100 follower examples:
+Below is an example of how we declare features with the `transform_features` method in the phone to SO-100 follower examples:
 
 ```python
     def transform_features(

docs/source/rtc.mdx

@@ -57,7 +57,7 @@ policy_cfg.rtc_config = RTCConfig(
 policy = PI0Policy.from_pretrained("lerobot/pi0_base", policy_cfg=policy_cfg, device="cuda")
 
 # Now use predict_action_chunk with RTC parameters
-inference_delay = 4  # How many steps of inference latency, this values should be calculated based on the inference latency of the policy
+inference_delay = 4  # How many steps of inference latency, this value should be calculated based on the inference latency of the policy
 
 # Initialize the action queue
 action_queue = ActionQueue(policy_cfg.rtc_config)
@@ -100,7 +100,7 @@ Typical values: 8-12 steps
 RTCConfig(execution_horizon=10)
 ```
 
-**`max_guidance_weight`**: How strongly to enforce consistency with the previous chunk. This is a hyperparameter that can be tuned to balance the smoothness of the transitions and the reactivity of the policy. For 10 steps flow matching (SmolVLA, Pi0, Pi0.5), a value of 10.0 is a optimal value.
+**`max_guidance_weight`**: How strongly to enforce consistency with the previous chunk. This is a hyperparameter that can be tuned to balance the smoothness of the transitions and the reactivity of the policy. For 10 steps flow matching (SmolVLA, Pi0, Pi0.5), a value of 10.0 is an optimal value.
 
 **`prefix_attention_schedule`**: How to weight consistency across the overlap region.
 

docs/source/smolvla.mdx

@@ -93,7 +93,7 @@ lerobot-train --help
 
 ## Evaluate the finetuned model and run it in real-time
 
-Similarly for when recording an episode, it is recommended that you are logged in to the HuggingFace Hub. You can follow the corresponding steps: [Record a dataset](./il_robots).
+Similarly for when recording an episode, it is recommended that you are logged in to the HuggingFace Hub. You can follow the corresponding steps: [Record a dataset](./il_robots#record-a-dataset).
 Once you are logged in, you can run inference in your setup by doing:
 
 ```bash

docs/source/using_dataset_tools.mdx

@@ -50,11 +50,11 @@ lerobot-edit-dataset \
 Divide a dataset into multiple subsets.
 
 ```bash
-# Split by fractions (e.g. 80% train, 20% test, 20% val)
+# Split by fractions (e.g. 60% train, 20% val, 20% test)
 lerobot-edit-dataset \
     --repo_id lerobot/pusht \
     --operation.type split \
-    --operation.splits '{"train": 0.8, "test": 0.2, "val": 0.2}'
+    --operation.splits '{"train": 0.6, "val": 0.2, "test": 0.2}'
 
 # Split by specific episode indices
 lerobot-edit-dataset \
@@ -252,6 +252,10 @@ lerobot-dataset-viz \
     --episode-index 0
 ```
 
+For a private or gated dataset, authenticate first with `hf auth login`, or set the
+`HF_TOKEN` environment variable. The Hub client then discovers the credential
+automatically; no token argument is needed.
+
 **From a local folder:**
 Add the `--root` option and set `--mode local`. For example, to search in `./my_local_data_dir/lerobot/pusht`: