create_zarr_from_csv

create_zarr_from_csv

Creates a Zarr-based replay buffer dataset from robot demonstration CSV files and associated images.

create_replay_buffer

create_replay_buffer(schema, datasets_paths)

Creates a Zarr-based replay buffer using a Hydra-instantiated dataset schema.

Parameters:

Name	Type	Description	Default
schema	CsvDatasetSchema	CsvDatasetSchema instance (instantiated by Hydra)	required
datasets_paths	list[str]	List of paths to episode CSV files	required

Source code in src/versatil/data/preprocessing/create_zarr_from_csv.py

def create_replay_buffer(schema: CsvDatasetSchema, datasets_paths: list[str]) -> None:
    """Creates a Zarr-based replay buffer using a Hydra-instantiated dataset schema.

    Args:
        schema: CsvDatasetSchema instance (instantiated by Hydra)
        datasets_paths: List of paths to episode CSV files
    """
    sorted_paths = sorted(datasets_paths, key=lambda x: int(Path(x).parent.name))
    episodes = _iter_csv_episodes(
        schema=schema,
        sorted_paths=sorted_paths,
    )

    create_zarr_replay_buffer(
        schema=schema,
        episodes=episodes,
        total_episodes=len(datasets_paths),
    )