[REQUEST] Enabling support for vision draft models.

[MoetasimR]

Problem

I am using a quantized version of pixtral large and I can't load the vision modules of a smaller variant. I cannot perform inference with images, I can only perform inference with text.
I imagine this will be a much needed feature as multimodal inference is always less performant than raw text.

Solution

Create a config for enabling this feature, I have a very strong feeling that this is low-hanging fruit.

Alternatives

No response

Explanation

I imagine this will be a much needed feature as multimodal inference is always less performant than raw text.

Examples

No response

Additional context

No response

Acknowledgements

• I have looked for similar requests before submitting this one.
• I understand that the developers have lives and my issue will be answered when possible.
• I understand the developers of this program are human, and I will make my requests politely.