Vision support for Qwen models.

[KevinDayve]

I think it going into Multimodal (image, video etc support) reasoning would be a great segue to the project you've built so far.

[NTT123]

Interestingly enough, visual chain-of-thought reasoning is the motivation behind this project for me. Though, I'm not sure I have enough time or commitment to pursue it right now. Maybe it's not that much additional work. In any case, if you'd like to work on it, I would very much welcome pull requests. 😄

[KevinDayve]

Sounds like a fun exercise :) The pertinent changes required would be in the qwen2_model.py I presume, where I would have to add vision tower support and process image embeddings.