Charles Gaillard

Hi, When you load the .h5 weights at inference time on the inference model, I don't really understand how you managed to save only the weights for the "inference model"...

### 🚀 The feature Run the demo with pytorch ### Motivation, pitch This would be great to be able to run the demo with pytorch ### Alternatives _No response_ ###...

feat: Add a loading bar when launching the demo

1

Add a loading bar to display instead of a blank page when loading the demo.

feature request: vision model support (Qwen2-VL 7b)

1

Hello there! Thanks for your amazing work. Do you plan in the near future to support vision model like Qwen2-VL ?