Fix bnb lora layers not setting active adapter #1294
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This now matches the behavior of the standard Linear lora layer.
younesbelkada
approved these changes
Dec 25, 2023
|
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. |
BenjaminBossan
approved these changes
Dec 28, 2023
There was a problem hiding this comment.
Thanks a lot for this bugfix, LGTM. @younesbelkada I think it means that there was no active adapter when using bnb layers.
|
Ok perfect, we can probably merge it then |
BenjaminBossan
added a commit
to BenjaminBossan/peft
that referenced
this pull request
Jan 11, 2024
Resolves huggingface#1345 See also huggingface#1294 for a similar (but incomplete) fix. This commit fixes the setting of the adapter name on a couple of quantized layers that was accidentally removed in huggingface#1106. This affects users who use a non-default adapter name when they want to train these layers.
BenjaminBossan
added a commit
that referenced
this pull request
Jan 12, 2024
Resolves #1345 See also #1294 for a similar (but incomplete) fix. This commit fixes the setting of the adapter name on a couple of quantized layers that was accidentally removed in #1106. This affects users who use a non-default adapter name when they want to train these layers. --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This appears to be a small oversight introduced with #1106. I was trying to load some loras onto LLMs using text-generation-webui, and noticed that the lora had no effect if the model was loaded with load_in_4bit. But non-quantized models still worked with the lora.
By adding this single line, it matches what is in the standard Linear lora layer. Now the quantized linear lora layers have the lora active by default.