lerobot-train --dataset.repo_id=staudi25/lego_pickup_maximus --job_name=pi05_so101_lego_pickup --policy.type=pi05 --policy.pretrained_path=lerobot/pi05_base --policy.device=cuda --wandb.enable=true --log_freq=1 --steps=7500 --save_freq=2500 --policy.repo_id=staudi25/pi05_lego_maximus_7500_no_aug --policy.dtype=bfloat16 --output_dir=outputs/train/pi05_lego_maximus --policy.compile_model=true --policy.gradient_checkpointing=true --policy.freeze_vision_encoder=false --policy.train_expert_only=false --batch_size=32

π₀.₅ is a Vision-Language-Action model with open-world generalization, from Physical Intelligence. The LeRobot implementation is adapted from their open source OpenPI repository.

π₀.₅ represents a significant evolution from π₀, developed by Physical Intelligence to address a big challenge in robotics: open-world generalization. While robots can perform impressive tasks in controlled environments, π₀.₅ is designed to generalize to entirely new environments and situations that were never seen during training.

For more details, see the Physical Intelligence π₀.₅ blog post.

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:

Prefix the dataset repo with eval_ and supply --policy.path pointing to a local or hub checkpoint.