Model Finetuning

#1
by zalim0zalima - opened

Hi there
Did you freeze the vision model weights during training?
And can I have any code related to finetuning the way you did?

Code is here: https://github.com/Li-Qingyun/mllm-mmrotate
the vision model is not frozen

Thanks.
As you already have worked on it, can you tell me
What will be models response if we limit output tokens to 10 or 20, Just for single object detection.

i do not exactly get you. i think 10 tokens is even tight for one HBB annotation. (two/three start tokens for three-beam-search, four box token and the other tokens for category name).

Your need to confirm your account before you can post a new comment.

Sign up or log in to comment