Disable kernels during calibration (and tracing) #1454
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
Signed-off-by: Kyle Sayers <[email protected]>
|
👋 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. |
dsikka
previously approved these changes
May 20, 2025
brian-dellabetta
previously approved these changes
May 21, 2025
Signed-off-by: Kyle Sayers <[email protected]>
brian-dellabetta
approved these changes
May 22, 2025
dsikka
approved these changes
May 22, 2025
aireilly
pushed a commit
to aireilly/llm-compressor
that referenced
this pull request
Jul 30, 2025
## Purpose ## * Guarantee that module hooks trigger by disabling kernel acceleration ## Background ## * As of `transformers>=4.52`, model forward functions may be overwritten with kernels * Kernel execution can be disabled using the config ([source](https://github.com/huggingface/transformers/blob/main/src/transformers/integrations/hub_kernels.py#L84)) * It seems that HF wants to continue enabling regular execution through this disabling feature. This gives us some faith that they will not [override the cls init function](https://github.com/huggingface/transformers/blob/main/src/transformers/integrations/hub_kernels.py#L78) in a way that inhibits execution with the standard forward definition * This only affects users who have the hf `kernels` library installed ## Changes ## * Implement `disable_hf_kernels` context and add to `calibration_forward_context` * Remove parenthesis around `calibration_forward_context` with in order to support python3.9 * Apply style to tests Signed-off-by: Kyle Sayers <[email protected]> Co-authored-by: Dipika Sikka <[email protected]>
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.
Purpose
Background
transformers>=4.52, model forward functions may be overwritten with kernelskernelslibrary installedChanges
disable_hf_kernelscontext and add tocalibration_forward_contextcalibration_forward_contextwith in order to support python3.9