Open StableLlama opened 4 weeks ago
It seems this is the ordinary meta-llama/Meta-Llama-3.1-8B
with the LoRA https://huggingface.co/spaces/fancyfeast/joy-caption-pre-alpha/blob/main/wpkklhc6/image_adapter.pt
on top
There's a fork of taggui that includes it now: https://github.com/doloreshaze337/taggui
There's a fork of taggui that includes it now: https://github.com/doloreshaze337/taggui
This doesn't actually seem to work? Following the instructions on that fork, it simply doesn't recognize the existence of the joycaption folder or the included imageadpter.pt. I have no idea why it doesn't, but it doesn't. Doesn't matter if you use the release or a git pull version, neither recognize the existence of it, so right now you can't use joycaption. Which sucks, because it's essentially required for tagging things for FLUX.
This doesn't actually seem to work?
When it's not working you should open up an issue there and not here.
This doesn't actually seem to work?
When it's not working you should open up an issue there and not here.
The reason I put it here is because this is a feature request, and I didn't want them to go 'well there's a fork so no need to add it to the main branch.' I think that this is something that should be added to the main branch support since it drastically improves tagging for FLUX models.
I'll look into adding it as soon as I have time (I'm not sure when that will be).
I'll look into adding it as soon as I have time (I'm not sure when that will be).
Glad to hear! Obviously, when you do it for free, one can't demand it be immediate. But I'm glad to hear it'll be added. Take your time though!
As seen at https://huggingface.co/spaces/fancyfeast/joy-caption-pre-alpha it is a model with superb captioning skills. It output was nearly creating the test image in Flux again. So it would be great when it could be used as a taggui autocaptioner.
The implementation could be copied from the spaces as linked above.