Join the community of Machine Learners and AI enthusiasts.
Sign Up
I don't have a lot of images, so I couldn't tell the difference between the optim and the ones that don't...🙀
They are twice the size.
In any case, it is certainly a Danbooru-type tagger with performance comparable to the WD14 series.
By the way, it worked even without Zero GPU space.
https://huggingface.co/spaces/John6666/danbooru-tags-transformer-v2-with-wd-tagger-b
from my tests it's kind of similar to wd14 but slightly worse for finding little details, it's probably because the limitation of cnns
btw the optim one is the optimizer state, it's not a model
im also pretraining a clip model with a classification head but it will most likely be worse than this one
