Handles the logic of setting up delta timestamps and image transforms before creating a dataset. Args: cfg (TrainPipelineConfig): A Train

Normalize a user-provided episode label to canonical lowercase values.

Discover and import third-party LeRobot plugins so they can register themselves. This function uses `importlib.metadata` to find packages in

Assign the correct ID and baud-rate to a single motor. This helper temporarily switches to the motor's current settings, disables torque, set

Serialize and write dataset statistics to their standard file path. Args: stats (dict): The statistics dictionary (can contain tensors/nu

Executes all processing steps and hooks in sequence. Args: transition: The initial `EnvTransition` object. Returns:

Get CAN ID for a motor.

Load configuration from local file or Hugging Face Hub. This method implements a super-simplified 3-way loading strategy: 1. **Local

Helper to save calibration data to the specified file. Args: fpath (Path | None): Optional path to save the calibration

Aggregates multiple LeRobot datasets into a single unified dataset. This is the main function that orchestrates the aggregation process by: 1

Aggregate stats from multiple compute_stats outputs into a single set of stats. The final stats will have the union of all data keys from each of

Apply this rotation to a set of vectors. This is equivalent to applying the rotation matrix to the vectors: self.as_matrix()

Convert a LeRobotDataset into a ReplayBuffer. Args: lerobot_dataset (LeRobotDataset): The dataset to convert.

Get the current action consumption index. Returns: int: Index of the next action to be consumed.

Convert a LeRobot features dictionary to a `datasets.Features` object. Args: features (dict): A LeRobot-style feature dictionary. Re

Get leftover original actions for RTC prev_chunk_left_over. These are the unconsumed actions from the current chunk, which will be us

Retrieves a policy class by its registered name. This function uses dynamic imports to avoid loading all policy classes into memory at o

Check if the package spec exists and grab its version to avoid importing a local directory. Args: pkg_name: The name of the package

Load dataset info metadata from its standard file path. Also converts shape lists to tuples for consistency. Args: local_dir (Path):

Merge multiple LeRobotDatasets into a single dataset. This is a wrapper around the aggregate_datasets functionality with a cleaner API. Args

Move the cursor up by a specified number of lines.

Predict a chunk of actions given environment observations.

Set the seed when entering a context, and restore the prior random state at exit. Example usage: ``` a = random.random() # produces som

Select action for inference/evaluation

Returns the state of the step (e.g., learned parameters, running means). Returns: A dictionary mapping state names to tensors.

Convert MM:SS or SS timestamp to seconds

Helper method to apply gradient checkpointing if enabled.

Decode a message and update the motor state cache.

Helper to save calibration data to the specified file. Args: fpath (Path | None): Optional path to save the calibration

Add a latency sample (seconds).

Tries to select automatically a torch device.

Calculate average FPS vs target

Compute comprehensive statistics for all features in an episode. Processes different data types appropriately: - Images/videos: Samples from

Compute dataset-level temporal proportions (priors) for each subtask. Implements SARM Paper Formula (1): ᾱ_k = (1/M) × Σ_i (L_{i,k} / T_i)

Convert dataset features to policy features. This function transforms the dataset's feature specification into a format that a policy can use

More info on ffmpeg arguments tuning on `benchmark/video/README.md`

Caps relative action target magnitude for safety.

Get a module's device by checking one of its parameters. Note: assumes that all parameters have the same device

Get file size on disk in megabytes. Args: file_path (Path): Path to the file.

Retrieve the current observation from the robot. Returns: RobotObservation: A flat dictionary representing the robot's c

Compute and return the statistics of the vectors processed so far. Args: quantiles: List of quantiles to compute (e.g., [0.01, 0.

Initializes a non-blocking keyboard listener for real-time user interaction. This function sets up a listener for specific keys (right arrow

Returns a client for the learner service. GRPC uses HTTP/2, which is a binary protocol and multiplexes requests over a single connection.

Loads normalization statistics from a flat state dictionary. This enables loading stats from safetensors files during from_pretraine

Constructs pre-processor and post-processor pipelines for the PI0 policy. The pre-processing pipeline prepares input data for the model by:

Constructs pre-processor and post-processor pipelines for the PI0 policy. The pre-processing pipeline prepares input data for the model by:

Constructs pre-processor and post-processor pipelines for the SmolVLA policy. The pre-processing pipeline prepares input data for the model

Modify a LeRobotDataset by adding and/or removing features in a single pass. This is the most efficient way to modify features, as it only copies

Predict a chunk of actions given environment observations.

Convert environment observation to LeRobot format observation. Args: observation: Dictionary of observation batches from a Gym vector envi

Interactively record the min/max encoder values of each motor. Move the joints by hand (with torque disabled) while the method streams live p

Resolve the final episode-success label from explicit and default values.

Centre each motor range around its current position. The function computes and writes a homing offset such that the present position becomes

Set random seed for all RNG sources to ensure reproducibility.

Start the controller and initialize resources.

Read the same value from multiple motors simultaneously.

Convert temporal proportions to cumulative breakpoints for normalization.

Writes `data` to a temporary JSON file and returns the file's path.

Context-manager that guarantees torque is re-enabled. This helper is useful to temporarily disable torque when configuring motors. E

Set training mode for augmentation decisions.