Skip to content

[Tests] add: test to check 8bit bnb quantized models work with lora loading. #10576

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

Merged
merged 4 commits into from
Jan 15, 2025

Conversation

sayakpaul
Copy link
Member

@sayakpaul sayakpaul commented Jan 14, 2025

What does this PR do?

To prevent from #10550. Only adds this test to 8bit. 4bit has issues and I will open a follow-up in a bit.

Slightly related: #10550.

@sayakpaul sayakpaul requested a review from DN6 January 14, 2025 11:48
sayakpaul and others added 2 commits January 15, 2025 08:54
Co-authored-by: Dhruv Nair <[email protected]>
@DN6
Copy link
Collaborator

DN6 commented Jan 15, 2025

Could consolidate by adding this to: #10578

@sayakpaul
Copy link
Member Author

Could consolidate by adding this to: #10578

I would prefer to do it differently because:

  1. Both are independent of one another. We never supported 8bit bnb models + LoRA loading (PEFT-side bug which recently got fixed).
  2. This is why the test being added in this PR has a requirement on the PEFT version.
  3. 4bit bnb + LoRA loading for Flux was supported in 0.31.0 but got broken. I think the scope of that is different from what we're doing in this PR.

@sayakpaul sayakpaul merged commit bba59fb into main Jan 15, 2025
11 checks passed
@sayakpaul sayakpaul deleted the 8bit-lora-loading branch January 15, 2025 07:35
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.

None yet

2 participants