-
Notifications
You must be signed in to change notification settings - Fork 5.6k
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
[SD3 LoRA Training] Fix errors when not training text encoders #8743
Conversation
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for the changes. Left some comments.
@asomoza let's apply the changes from the review comments then. Before merging, let's also perhaps do two runs -- with and without text encoder training. |
@sayakpaul @asomoza btw just tried an experiment only with the change of |
Co-authored-by: Linoy Tsaban <[email protected]>
@linoytsaban and I took the liberty of fixing some things. Hope that is fine. I have run some basic training on my end and things seemed to be working. It'd be nice if you could verify that as well (Linoy, Alvaro, both). @neuron-party could you also check if this PR resolves the problems you were facing? |
@linoytsaban @sayakpaul I tested it with just |
Thanks! We can merge afterward then. Let us know. |
Thanks @asomoza! looks good 🔥 |
Thanks all :) |
* fix * fix things. Co-authored-by: Linoy Tsaban <[email protected]> * remove patch * apply suggestions --------- Co-authored-by: Linoy Tsaban <[email protected]> Co-authored-by: sayakpaul <[email protected]> Co-authored-by: Linoy Tsaban <[email protected]>
What does this PR do?
Adds some validations to the SD3 dreambooth lora training when not training the text encoders.
Fixes the following errors:
train command:
and
Note: I only added validations when the variables are None, didn't check the logic or the result of the training.
Who can review?
@sayakpaul @linoytsaban