search

luogen1996 / LLaVA-HR

LLaVA-HR: High-Resolution Large Language-Vision Assistant
Apache License 2.0
202 stars 9 forks source link

How to load convnext model using local bin file #7

Open yuecao0119 opened 5 months ago

yuecao0119 commented 5 months ago

Hello, thanks for your work.

But I would like to ask if I want the convnext model to use local files instead of downloading them every time. How should this be done? I tried modifying the code, but to no avail.

Could you please give me some guidance? Thank you very much.

aoji0606 commented 5 months ago

I have the same problem

gapjialin commented 5 months ago

I have the same problem

luogen1996 commented 5 months ago

Download the weights from here, then change the name --vision_tower_slow convnext_large_mlp.clip_laion2b_ft_320 to your download path, e.g., ./weights/convnext_large_mlp.clip_laion2b_ft_320 .

aoji0606 commented 5 months ago

I did it exactly the way you said, but it didn't work image image

luogen1996 commented 5 months ago

This is because your network cannot connect to the huggingface server. As far as I know, even loading weights locally may still require access to the huggingface server. I recommend you to use a VPN such as pigcha.

chuangzhidan commented 5 months ago

clip_laion2b_ft_320

which model/weights should we download exactly?because there are so many choices.