bugfix kv cache quantization with ignored layers #1312
Merged
Conversation
This file contains hidden or 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
|
👋 Hi! Thank you for contributing to llm-compressor. Please add the ready label when the PR is ready for review. Note: This is required to complete the testing suite, please only add the label once the PR is code complete and local testing has been performed. |
kylesayrs
reviewed
Apr 1, 2025
brian-dellabetta
force-pushed
the
bdellabe/kv_cache_ignore_layer_bugfix
branch
from
9d7b26e to
8bc85a9
Compare
@kylesayrs I originally started with that, but since this is always incremented it ends up being a smaller footprint and we can still get by with easy access with |
rahul-tuli
approved these changes
Apr 2, 2025
kylesayrs
approved these changes
May 5, 2025
Signed-off-by: Brian Dellabetta <[email protected]>
Signed-off-by: Brian Dellabetta <[email protected]>
brian-dellabetta
force-pushed
the
bdellabe/kv_cache_ignore_layer_bugfix
branch
from
8bc85a9 to
aad85a3
Compare
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Outdated suggestions cannot be applied.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.
SUMMARY:
Users may ignore some layers in configuration, meaning list indices may not correspond to a given layer index when values are being appended to a list for each layer. We must account for that case by padding list so that index of lists always corresponds to layer_idx
Resolves #1295 . I can confirm i see the error original poster is reporting, using their recipe with
meta-llama/Meta-Llama-3-8B-Instruct, and that this PR resolves it. I am able to runlm_evalon the quantized model. I will move this to ready for review to get feedback.TEST PLAN:
Add doctests to unit test new helper function, not sure if these get tested though.