Openvla 7b Prismatic

🚀 OpenVLA 7B (Prismatic-Compatible Version)

This model is a variant of the OpenVLA 7B model, with its checkpoint formatted to be compatible with the training script from the original Prismatic VLMs project codebase. The OpenVLA team developed the OpenVLA model based on this codebase. For more details about the OpenVLA 7B model, visit: https://huggingface.co/openvla/openvla-7b.

This Prismatic-compatible checkpoint is useful for those who want to fully fine-tune OpenVLA (all 7.5 billion parameters) using native PyTorch Fully Sharded Data Parallel (FSDP) with the Prismatic VLMs training script. If you prefer to perform Parameter-Efficient Fine-Tuning via LoRA, you can use the OpenVLA checkpoint linked above, which is compatible with the Hugging Face transformers library. We recommend fine-tuning via LoRA if you lack the necessary computing resources to fully fine-tune a 7B-parameter model (e.g., multiple A100/H100 GPUs).

🚀 Quick Start

See the OpenVLA GitHub README for instructions on how to use this checkpoint for full fine-tuning.

📚 Documentation

Model Information

📄 License

This project is licensed under the MIT license.

📚 Citation

BibTeX:

@article{kim24openvla,
    title={OpenVLA: An Open-Source Vision-Language-Action Model},
    author={{Moo Jin} Kim and Karl Pertsch and Siddharth Karamcheti and Ted Xiao and Ashwin Balakrishna and Suraj Nair and Rafael Rafailov and Ethan Foster and Grace Lam and Pannag Sanketi and Quan Vuong and Thomas Kollar and Benjamin Burchfiel and Russ Tedrake and Dorsa Sadigh and Sergey Levine and Percy Liang and Chelsea Finn},
    journal = {arXiv preprint arXiv:2406.09246},
    year={2024}
}