iloveai111/frankamode

Franka Pi0.5 LoRA checkpoint

OpenPI Pi0.5 LoRA fine-tuned checkpoint for Franka FR3.

Training

• Config: pi05_franka_test_low_mem
• Base checkpoint: gs://openpi-assets/checkpoints/pi05_libero/params
• Dataset: iloveai111/franka_test
• Step: 999
• Fine-tuning method: LoRA
• Action format: Franka 9D, joint actions are delta, fingers are absolute

Usage

python scripts/serve_policy.py \
  --policy.config=pi05_franka_test_low_mem \
  --policy.dir=hf://iloveai111/frankamode \
  --port=8000

If hf:// is not supported by your local OpenPI version, download this repository and pass the local directory as --policy.dir.