Any4Robot

any4robot provides dataset conversion and management utilities for training VLA / WM / VAM robot control models. Data preparation is crucial for robot model development. Any4Robot provides a standardized, verified way to produce the right data so you can focus on model design.

Install

pip install any4robot

Quick start (LeRobot editor)

from any4robot import LeRobotDatasetEditor

editor = LeRobotDatasetEditor("/path/to/lerobot-dataset")
editor.list_episodes(0, 5)
editor.delete_episode(12, dry_run=True)

Combine datasets

from any4robot.combine import merge_lerobot_datasets

merge_lerobot_datasets(
    ["/path/to/dataset_a", "/path/to/dataset_b"],
    "/path/to/output_dataset",
)

Quality check (LeRobot v2.1 layout)

from any4robot.qualitycheck import validate_lerobot_v2_1

result = validate_lerobot_v2_1("/path/to/lerobot-dataset")
print(result)

Modules

• any4robot.editing: LeRobot dataset editing helpers (delete episode, copy episode, list episodes)
• any4robot.combine: Merge multiple LeRobot datasets into a single dataset
• any4robot.qualitycheck: Validate datasets (LeRobot v2.1 layout)
• any4robot.conversion: Placeholder for future dataset conversion pipelines

Notes

• The LeRobot editor code is copied from vendor/lero and organized under any4robot/lerobot_editor.
• Optional GUI dependencies are available with pip install any4robot[gui].