[Request] Quantized Version

[QwertyJack]

Thanks for your great work!

Is there any consideration for an official quantized version just like what Qwen does? GPTQ or AWQ are preferred. FYI: https://hf-mirror.com/collections/Qwen/qwen2-vl-66cee7455501d7126940800d

[rubentorresbonet]

+1, at least for 7B-DPO please

[Taoran-Lu]

Thank you for your interest! We are currently discussing solutions regarding model quantization. We will keep the community updated as we make progress. Please feel free to continue sharing your suggestions and feedback.