LLaVA
LLaVA copied to clipboard
[Question]
Question
I wanted to fine-tune LLaVA on my dataset for my use case, but I had only 553 images in total. When I fine-tuned, merged the weights and then checked the inference of the fine-tuned model it was not at all good.
Any suggestions @haotian-liu as how I can improve the inference or where did I make the mistake?