LoRA: dequantize model #811
Draft
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
Andrei-Aksionov
requested review from
awaelchli,
carmocca and
lantiga
as code owners
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Outdated suggestions cannot be applied.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.
Hi there 馃憢
This PR adds functionality to dequantize the model that was initially quantized, in case anyone wants to use these weights elsewhere.
The logic is to dequantize to
float16dtype, that is hardcoded inquant_stateand then cast to the compute dtype, the one that was used for trainable LoRA parameters and data inputs.This is exactly how it's done in BNB.
Note: the PR is not yet finalized as, for completeness, I also want to add dequantization for
bnb.int8.