Action Chunking with Transformers (ACT) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.
This policy has been trained and pushed to the Hub using
LeRobot.
See the full documentation at
LeRobot Docs.
For a complete walkthrough, see the
training guide.
Below is the short version on how to train and run inference/eval:
1lerobot-train \
2 --dataset.repo_id=${HF_USER}/<dataset> \
3 --policy.type=act \
4 --output_dir=outputs/train/<desired_policy_repo_id> \
5 --job_name=lerobot_training \
6 --policy.device=cuda \
7 --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
8 --wandb.enable=true
1lerobot-record \
2 --robot.type=so100_follower \
3 --dataset.repo_id=<hf_user>/eval_<dataset> \
4 --policy.path=<hf_user>/<desired_policy_repo_id> \
5 --episodes=10
