lerobot/pi0fast_base

π0+FAST: Efficient Action Tokenization for Vision-Language-Action Models

Paper Jax code

Designed by Physical Intelligence. Ported from Jax by Hugging Face.

Example of finetuning the pi0+FAST pretrained model (pi0_fast_base in openpi):

python lerobot/scripts/train.py \
--policy.path=lerobot/pi0fast_base \
--dataset.repo_id=danaaubakirova/koch_test

Example of training the pi0+FAST neural network with from scratch:

python lerobot/scripts/train.py \
--policy.type=pi0fast \
--dataset.repo_id=danaaubakirova/koch_test

Example of using the pi0 pretrained model outside LeRobot training framework:

policy = PI0FASTPolicy.from_pretrained("lerobot/pi0fast_base")