search

microsoft / LLaVA-Med

Large Language-and-Vision Assistant for Biomedicine, built towards multimodal GPT-4 level capabilities.
Other
1.29k stars 148 forks source link

checkpoints #3

Open zuwenqiang opened 1 year ago

zuwenqiang commented 1 year ago

Hi, It is really a great work! May I ask if you will release the checkpoint for llava-med later? Or how can we fine-tune based on llava? Is there any dataset available?

latent-variable commented 1 year ago

I am doing research on the capabilities of multimodel models for applications in podiatry and would love to have access to the checkpoint.

jlamperez commented 1 year ago

Where is the model?

hellocym commented 8 months ago

Hi, It is really a great work! May I ask if you will release the checkpoint for llava-med later? Or how can we fine-tune based on llava? Is there any dataset available?

I found a repo called LLaVA-Med-preview and found a checkpoint inside it. I managed to run an inference on VQA-Rad with it, but I couldn't reproduce the result (0 epochs on VQA-Rad, with an accuracy of 0.61) in the paper. :( I think it could be something wrong in my evaluation code, or I used inappropriate parameters during inference.

ChunyuanLI commented 8 months ago

LLaVA-Med is finally released after 5-month intensive discussions to go through Microsoft release process. Thanks for your patience.

SinanAkkoyun commented 8 months ago

Thank you so so much! @ChunyuanLI