Added detailed captioning, increase `max_new_tokens` and fix escape character

#8
by merve HF staff - opened
No description provided.

Sounds good, I think the DETAILED_CAPTION_PROMPT should also have \n appended? Not sure about this one, cc @mtensor

I also didn't put newline at the end of VQA prompts, and this was a VQA prompt more than captioning so I didn't

it looks like this:
Screenshot 2023-10-20 at 17.30.12.png

I moved these changes to new PR #12, which also includes the screenshot text location feature. I'd suggest we keep the current prompts given that they work, but happy to tweak them with feedback from @mtensor :)

pcuenq changed pull request status to closed

Sign up or log in to comment