---
library_name: transformers
license: apache-2.0
base_model: google/siglip2-so400m-patch16-384
tags:
  - vision
  - image-encoder
  - siglip
  - siglip2
---

# SigLIP 2 So400m Patch16 384 (MambaMia Compatible)

This is a modified version of [google/siglip2-so400m-patch16-384](https://huggingface.co/google/siglip2-so400m-patch16-384) with preprocessing configurations adjusted for compatibility with [MambaMia](https://github.com/naver-ai/mambamia).

## Modifications

The only modification from the original model is in `preprocessor_config.json` to ensure compatibility with the MambaMia framework. **The model weights remain unchanged.**

## Original Model

- **Base Model:** [google/siglip2-so400m-patch16-384](https://huggingface.co/google/siglip2-so400m-patch16-384)
- **Developed by:** Google
- **Model type:** Vision Encoder (SigLIP 2)
- **License:** Apache 2.0

## Usage

This model is intended to be used as a vision encoder within the MambaMia framework. Please refer to the [MambaMia repository](https://github.com/naver-ai/mambamia) for detailed usage instructions.

## Citation

If you use this model, please cite the original SigLIP 2 paper and MambaMia:

```bibtex
@article{tschannen2025siglip,
  title={SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features},
  author={Tschannen, Michael and others},
  journal={arXiv preprint},
  year={2025}
}
```

## License

This model is released under the [Apache 2.0 License](https://www.apache.org/licenses/LICENSE-2.0), following the original model's license.