feat(dataset): add subtask support (#2860)

* add subtask

* remove folder

* add docs

* update doc

* add testing

* update test

* update constant naming + doc

* more docs
This commit is contained in:
Jade Choghari
2026-01-30 10:29:37 -08:00
committed by GitHub
parent 5c6182176f
commit b18cef2e26
9 changed files with 1003 additions and 2 deletions
+9
View File
@@ -60,6 +60,7 @@ VIDEO_DIR = "videos"
CHUNK_FILE_PATTERN = "chunk-{chunk_index:03d}/file-{file_index:03d}"
DEFAULT_TASKS_PATH = "meta/tasks.parquet"
DEFAULT_SUBTASKS_PATH = "meta/subtasks.parquet"
DEFAULT_EPISODES_PATH = EPISODES_DIR + "/" + CHUNK_FILE_PATTERN + ".parquet"
DEFAULT_DATA_PATH = DATA_DIR + "/" + CHUNK_FILE_PATTERN + ".parquet"
DEFAULT_VIDEO_PATH = VIDEO_DIR + "/{video_key}/" + CHUNK_FILE_PATTERN + ".mp4"
@@ -353,6 +354,14 @@ def load_tasks(local_dir: Path) -> pandas.DataFrame:
return tasks
def load_subtasks(local_dir: Path) -> pandas.DataFrame | None:
"""Load subtasks from subtasks.parquet if it exists."""
subtasks_path = local_dir / DEFAULT_SUBTASKS_PATH
if subtasks_path.exists():
return pd.read_parquet(subtasks_path)
return None
def write_episodes(episodes: Dataset, local_dir: Path) -> None:
"""Write episode metadata to a parquet file in the LeRobot v3.0 format.
This function writes episode-level metadata to a single parquet file.