Vikhr 2 VL 2b Instruct Experimental

🚀 Vikhr-2-VL-2b-Instruct-experimental

Vikhr-2-VL-2b-Instruct-experimental is a compact VLM model based on Qwen2-VL-2B-Instruct. It is trained on a translated LLAVA-150K dataset and fine-tuned specifically for Russian language processing. The fine-tuned model is experimental and may not always perform as expected (especially in OCR). For feedback, please use Vikhr Models.

🚀 Quick Start

Try now:

✨ Features

• Compact Model: Based on Qwen2-VL-2B-Instruct, it offers a lightweight solution for vision-language tasks.
• Multilingual Support: Supports both Russian and English, making it suitable for a wider range of users.
• Experimental Fine-tuning: Fine-tuned on a translated dataset for better performance in Russian language processing.

📄 License

This project is licensed under the Apache-2.0 License.

📚 Documentation

Authors

• Nikolay Kompanets, LakoMoor, Vikhr Team
• Sergey Bratchikov, NlpWonder, Vikhr Team
• Konstantin Korolev, underground, Vikhr Team
• Aleksandr Nikolich, Vikhr Team

@inproceedings{nikolich2024vikhr,
  title={Vikhr: Constructing a State-of-the-art Bilingual Open-Source Instruction-Following Large Language Model for {Russian}},
  author={Aleksandr Nikolich and Konstantin Korolev and Sergei Bratchikov and Nikolay Kompanets and Igor Kiselev and Artem Shelmanov },
  booktitle = {Proceedings of the 4rd Workshop on Multilingual Representation Learning (MRL) @ EMNLP-2024},
  year={2024},
  publisher = {Association for Computational Linguistics},
  url={https://arxiv.org/pdf/2405.13929}
}

📦 Information Table

⚠️ Important Note

The fine-tuned model is experimental and may not always perform as expected, especially in OCR tasks.