Minivla Vq Bridge Prismatic

🚀 MiniVLA 1B VQ Trained on Bridge V2 (Prismatic-Compatible Version)

This checkpoint is compatible with the training script from the original Prismatic VLMs project. It offers options for fine - tuning MiniVLA, either fully or via Parameter - Efficient Fine - Tuning.

🚀 Quick Start

This Prismatic - compatible checkpoint is in a format that is compatible with the training script from the original Prismatic VLMs project codebase. The OpenVLA team built on this codebase to develop the OpenVLA model.

If you want to fully fine - tune MiniVLA (all 1 billion parameters) via native PyTorch Fully Sharded Data Parallel (FSDP) using the Prismatic VLMs training script, this checkpoint may be useful. However, if you prefer to do Parameter - Efficient Fine - Tuning via LoRA, you can use the MiniVLA checkpoint linked above, which is compatible with the Hugging Face transformers library. We recommend fine - tuning via LoRA if you don't have enough computing resources to fully fine - tune a 1B - parameter model (e.g., multiple A100/H100 GPUs).

📚 Documentation

Usage Instructions

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

📄 License

The license for this project is MIT.

📚 Citation

BibTeX:

@article{belkhale24minivla,
    title={MiniVLA: A Better VLA with a Smaller Footprint},
    author={Suneel Belkhale and Dorsa Sadigh},
    url={https://github.com/Stanford-ILIAD/openvla-mini}
    year={2024}
}