Token not working in downloaded model but works on hugging face
So i trained a model and downloaded the model.ckpt and added the yaml file in automatic1111 i noticed that my token would not work in automatic1111 yet it works fine i try it on hugging face in the Hosted inference API in text to image.... what am i doing wrong ?
isn't model the ckpt the right file to download? is there anything else that i need to download to get it running locally ?
Hey!
I'm having the exact same problem
I have used the v2-1-512 model as a base and it works really well in the Hosted interface API as well as in the Google colab for Inference notebook (https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_dreambooth_inference.ipynb)
But as soon as I want to use the ckpt (together with the standard 2.0 yaml) in the automatic1111 copy that I have running on my local workstation, the prompts stop working and I just get some random results.
Does anyone know why it doesn't work?
i assume this version of dreambooth is only created diffusers ? not generating a ckpt model ? is there any way to do the conversion ? the results are great but my main use case is to use this locally
The model.ckpt you mentioned before should (theoretically) work.
Followed this workflow: https://twitter.com/angrypenguinPNG/status/1599611909623332864?s=20&t=AbseaL4v_NScij5xplyvYA
I've seen a video with a description on YT (https://www.youtube.com/watch?v=_e5ymV4zY3w) on how to convert manually, but I haven't tried it yet myself.
followed that as well but it didn't work for me, i also tried the method from the video but it seems that it requires linux so didn't work for me, i also tried the google collab to just point to my files on google drive but that didn't work for me either :/
did you by any chance also change the diffusers version in the requiremetns.txt ? could that be the reason why its not working ? some funkiness with the diffusers version ?
I have the same problem PLEASE SOMEONE HELPS
Hi everyone, great points. The convert_diffusers_to_original_stable_diffusion.py this Space uses was not fully compatible with 2.1
The script was updated Today for full compatibility and I will update it here accordingly for future trains.
Those who already have a working diffusers model but with a broken *.ckpt can convert it via:
- The convert_diffusers_to_original_stable_diffusion.py script
- Or this Colab that I just made quickly https://colab.research.google.com/drive/1jmUP5ZhM4M63l9UFoiAVgUyjfudO9lmK for you
update it actually works, my bad i had an old model in my downloads folder i assumed that's the one :)
hey @multimodalart !
Thank you so much for the updated script! I just tried to convert my models with your colab and now it works like a charm!