cjgogo/pi0-aloha-robotwin-lora_place_container_plate

Pi0 ALOHA RobotWin LoRA Checkpoint

This is a Pi0 base model fine-tuned with LoRA on the ALOHA RobotWin dataset.

Model Details

• Base Model: Pi0 Base
• Fine-tuning Method: LoRA
• Dataset: demo_clean
• Training Steps: 30000
• Checkpoint Size: ~15GB

Checkpoint Structure

30000/
├── assets/          # Model assets
├── params/          # Model parameters
├── train_state/     # Training state
└── _CHECKPOINT_METADATA

Usage

from pi0 import load_checkpoint

checkpoint_path = "path/to/30000"
model = load_checkpoint(checkpoint_path)

Training Info

• Checkpoint directory: /data/mayue/cjy/RoboTwin/policy/pi0_place/checkpoints/pi0_base_aloha_robotwin_lora/demo_clean/30000
• Training completed at step 30000

Citation

If you use this checkpoint, please cite:

@misc{pi0-aloha-robotwin,
  title={Pi0 ALOHA RobotWin LoRA Checkpoint},
  author={Your Name},
  year={2025},
  howpublished={\url{https://huggingface.co/cjgogo/pi0-aloha-robotwin-lora_place_container_plate}}
}