Llava 1.6?

[ElliottDyson]

Any chance we could see a variant of each produced with the Llava 1.6 architecture? Thanks

[ElliottDyson]

@mmaaz60 Is there any way I can help you or are you just waiting on compute availablity/grants?

[mmaaz60]

Hi @ElliottDyson,

Thank you for your interest in our work. Nowadays, its been very tight schedule and is difficult to spare some time for any extra experiment. However, LLaVA 1.6 (LLaVA Next) preprocessor is implemented in transformers at https://github.com/huggingface/transformers/blob/e0c3cee17085914bbe505c159beeb8ae39bc37dd/src/transformers/models/llava_next/image_processing_llava_next.py#L117 and can be integrated and used with LLaMA-3-V and Phi-3-V.

We will welcome and appreciate any pull request implementing it. Thank You

[ElliottDyson]

Hi @ElliottDyson,

Thank you for your interest in our work. Nowadays, its been very tight schedule and is difficult to spare some time for any extra experiment. However, LLaVA 1.6 (LLaVA Next) preprocessor is implemented in transformers at https://github.com/huggingface/transformers/blob/e0c3cee17085914bbe505c159beeb8ae39bc37dd/src/transformers/models/llava_next/image_processing_llava_next.py#L117 and can be integrated and used with LLaMA-3-V and Phi-3-V.

We will welcome and appreciate any pull request implementing it. Thank You

So it is, thank you. I'll start looking into it in a week's time from now

[gesen2egee]

Is there any news? : D Also, can this integration be achieved by just modifying the code?