Fix QLoRA weights and bias initialisation #833
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
anwai98
reviewed
Jan 15, 2025
There was a problem hiding this comment.
Hi @caroteu,
I left some comments below. In general, I feel using strict=False is a bit tricky (and error-prone) to set it for loading model checkpoints. We should discuss the issues and find an elegant way to take care of this!
|
@anwai98 |
anwai98
reviewed
Jan 20, 2025
anwai98
reviewed
Jan 20, 2025
Does the inference for QLoRA-finetuned models properly!
|
I added docstrings to our new export function (and thanks to @caroteu for making an elegant solution out of this). This is GTG from my side. PS. @caroteu @constantinpape It would be nice to go over this and see if everything makes sense now! |
constantinpape
approved these changes
Jan 21, 2025
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Outdated suggestions cannot be applied.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.
@anwai98 This includes the QLoRA initialisation changes and this also works for inference now. The changes that made the inference work are a bit hacky though, so we should think about a better/safer way to do this.