ColPali
Safetensors
English
vidore

how improve performance

#8
by marcoaleixo - opened

while trying on colab I'm far from having 0.39s per image.
I'm basically running your example ( thanks ) https://github.com/illuin-tech/colpali/blob/main/scripts/infer/run_inference_with_python.py#L14 but using a T4.
I'm seeing almost 15/20s per page, any clues?
and I noticed that you mentioned that you used a nvidia L4, I will try with that one.

thanks

Sign up or log in to comment