Skip to content

Conversation

@kyleengan
Copy link

@kyleengan kyleengan commented Aug 2, 2025

Steps to reproduce bug this PR is (theoretically) fixing (issue link: #117):

  1. Start Forge Classic with --persistent-patches feature flag turned on
  2. Load a model, add a prompt containing usage of at least one Lora
  3. Set a non-random Seed for easier observation
  4. Hit "Generate"
  5. Significantly alter a Lora weight in the prompt (this makes it easier to check for changes to the image)
  6. Hit "Generate"
  7. [BUG] Observe the utter lack of change in the image
  8. PRESUMED CAUSE: the persistent patches feature (correctly) prevented an unpatch after the first Generate, so the updated Lora weights aren't correctly patched in
  9. Hit "Generate" again (no changes to prompt/loras!)
  10. Observe the image change correctly according to the Lora weight alteration in step 5.

Persistent Patches is pretty much THE reason I switched to using Forge Classic, but I noticed that I had to hit Generate an extra time each time I fiddled with Lora weights. This bugged me enough to submit a PR with a proposed fix.

Thanks!

NOTE: I'm not a Python dev at all, so please let me know if any changes are preferred to the proposed implementation (or, if you don't wanna wait on me, I won't be offended if you fix the PR up how you like prior to inclusion, assuming you want to merge this :) )

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