How to prompt MiaoshouAI/Florence-2-base-PromptGen

[asdrabael]

If you do no prompt at all, it just does booru tags. It has the ability to do detailed captions, but if you try to use anything it says:

This model only supports the following prompts: "<GENERATE_PROMPT>", "", "<DETAILED_CAPTION>", and "<MORE_DETAILED_CAPTION>". The default prompt is "<GENERATE_PROMPT>".

I've tried "<DETAILED_CAPTION>", prompt="<DETAILED_CAPTION>", and every way I can think of but I can't get a response except for that. Am I missing something obvious? I'm sure it's something simple.

[jhc13]

Have you tried using "<DETAILED_CAPTION>" as the prompt?

[asdrabael]

Yes, i wrote that in my post but the html broke it. I fixed it but ended up having to open the florence2 python code in the auto-captioning folder and changing the default in the python code because no prompts worked.