Reloading the text model fails, saying Tokenizer in Transformers format was not found
I'm using the VRAM option since my text model is very large. It seems an image gets created, but then it has trouble reloading the model:
text-generation-webui loads the model fine, but then it doesn't seem to be correctly reloading it after stable_diffusion unloads it to run the image generator.
It looks like it is trying to load "None" for some reason...?
I fixed this in my fork: https://github.com/Trojaner/text-generation-webui-stable_diffusion/commit/ab6d2a15ef78bc9a87793481a5cf7db1a63a9982
I checked out your fork, can you do a PR once your changes are finalized?
Thanks!
I didn't do a great job of isolating just these changes, as I made some other improvements on my fork. It might be easier to just cherry pick these two commits to fix this bug:
https://github.com/Trojaner/text-generation-webui-stable_diffusion/commit/ab6d2a15ef78bc9a87793481a5cf7db1a63a9982
https://github.com/Trojaner/text-generation-webui-stable_diffusion/commit/fb2a9aac52f2492540ecefdf813dd2df12661408
Its merged now, thanks 👍