Patch lora kernels post model load #2345
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
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Outdated suggestions cannot be applied.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.
Description
This PR changes the way in which we're patching the attention class with custom QKV / output projections code in order to apply our custom Triton kernels / autograd functions.
After model load, we have knowledge of which attention class is used in the model. This is not the case pre-model load, where we would have to maintain a data structure mapping
model_types to their attention implementation(s).Motivation and Context
We can be more flexible / dynamic and support patching more models with the LoRA optims.
How has this been tested?
TODO: comprehensive tests on remaining (or high-value?) model types.