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

Add Option for Pyramid Noise to SDXL train script #5591

Closed
wants to merge 4 commits into from

Conversation

nbardy
Copy link
Contributor

@nbardy nbardy commented Oct 30, 2023

This changes the offset_nose arg to be a noise_type arg.

Currently allows 3 options between:

  • Standard white noise
  • Pyramid Noise
  • Offset Noise

@sayakpaul
@patrickvonplaten

@sayakpaul
Copy link
Member

Thank you for your PR! But we're currently holding off a bit from merging new changes into our official training examples to keep them minimal.

I welcome you to add a note in the corresponding READMEs of the examples:

  • Brief commentary
  • Pseudo code with source
  • Results expected

@nbardy
Copy link
Contributor Author

nbardy commented Oct 31, 2023

Will close this then.

Definitely feeling some of the pain of trying to fork this script given how complicated it has gotten.

Would be nice to have a train_sdxl_simple.py in the repo that is barebones and way less lines of code.

@sayakpaul Not sure what you mean by adding to the readme? Adding chunks of code to the readme? Don't quiet understand the goal/purpose of that.

@nbardy nbardy closed this Oct 31, 2023
@sayakpaul
Copy link
Member

Thanks for understanding!

@sayakpaul Not sure what you mean by adding to the readme? Adding chunks of code to the readme? Don't quiet understand the goal/purpose of that.

I mean adding a brief section about it and supplementing that with pseudocode if necessary.

Would be nice to have a train_sdxl_simple.py in the repo that is barebones and way less lines of code.

SDXL itself has some extra components that need working around. If we want to follow the guidelines as laid out here, this is the reasonable version we could come up with. If you have improvement idea, let's maybe discuss in an issue.

Cc: @patrickvonplaten

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

Successfully merging this pull request may close these issues.

2 participants