Skip to content
This repository was archived by the owner on Oct 9, 2025. It is now read-only.

Conversation

@matt-bornstein
Copy link

This is a new tab to run a bunch of SDXL fine tunes on a single prompt.

It pulls models from https://replicate.com/collections/sdxl-fine-tunes.

The hard-ish part is deciding how to prompt each model. I hard-coded templates for a bunch of them in lib/sdxlfinetunes.js. If the app encounters a model not in that set, it uses llama2 to guess at the format from a single example. The logic for this is in pages/api/predictions/sdxlfinetunes.js.

This could be improved in various ways:

  • Store the templates in supabase instead of hard coded
  • Add more templates for the rest of the models
  • Pull more example prompts for each model before guessing (few shot vs single shot)
  • Improve llama2 prompt in some other way

I'm sure I broke a bunch of conventions etc, so happy to change anything

@vercel
Copy link

vercel bot commented Sep 11, 2023

@matt-bornstein is attempting to deploy a commit to the Replicate Team on Vercel.

A member of the Team first needs to authorize it.

@vercel
Copy link

vercel bot commented Oct 19, 2023

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name Status Preview Comments Updated (UTC)
zoo ✅ Ready (Inspect) Visit Preview 💬 Add feedback Oct 19, 2023 5:58pm

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant