Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Plan for integration of SDXL? It seems to require more work than changing the MODEL_NAME #73

Open
asrlhhh opened this issue Aug 7, 2023 · 6 comments

Comments

@asrlhhh
Copy link

asrlhhh commented Aug 7, 2023

I tried editing the training command like export MODEL_NAME="stabilityai/stable-diffusion-xl-base-1.0" but it seems the SDXL model has some structural change on weight files from the previous versions and require some changes in the training file as well. For me, the errors start when it fails to load text_conder. I traced down to the text_encoder folder and it appears that the file structures of text_encoder of SDXL is different than that of 2.1 and 1.4. I guess this is true for other parts of the model loading as well. I am not sure if the team has a plan for that

@nupurkmr9
Copy link
Collaborator

Hi, thanks a lot for your interest in our work.
I aim to try custom diffusion with SDXL, but I don't have a fixed timeline by when I will be able to update the repo with that. I will comment here once I have updated our method with SDXL support.

@Sandeep-Narahari
Copy link

Sandeep-Narahari commented Sep 12, 2023

Thanks for the amazing work team!
May I know any update on SDXL support, I am also waiting for the SDXL support from last month

@Oocean35
Copy link

Hi @nupurkmr9, Thanks for the amazing work !
Any update on SDXL support ?
Thanks a lot

@nupurkmr9
Copy link
Collaborator

Hi, I have updated the code with SDXL support. Let me know if you face any issues.
Thanks!

@shyammarjit
Copy link

@nupurkmr9 Please provide the bash scripts to run SDXL pipeline.

@nrsz00011
Copy link

nrsz00011 commented Mar 24, 2024

Hi, I have updated the code with SDXL support. Let me know if you face any issues. Thanks!

Great work! Thank you
It seems the current script (diffusers_training_sdxl.py) is not supporting SDXL for multi-concept.
Line 1113 only handles the args of the single concept
and in the training loop, line 1239, also only handles the single concept token with ignoring the multi-concept tokens.

I would really appreciate it if the support added for multi-concept SDXL

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

6 participants