Spaces:

Duplicated from  MaziyarPanahi/llava-llama-3-8b

MaziyarPanahi

/

Phi-3-vision-128k

like 44

Running on Zero

well flash_attn really requires lots of stuff including cuda related things, so i'm going to close this pr
good luck @MaziyarPanahi

Thank you @not-lain

I think it should be either Dockerfile or use subprocess inside the app.py for flash_attn, or else as you mentioned it will fail.