fix: enabling block-by-block evaluation for granite-3.x-models#165
Merged
chichun-charlie-liu merged 3 commits intofoundation-model-stack:mainfrom Jul 24, 2025
Merged
Conversation
bayo-ibm
requested review from
BrandonGroth,
andrea-fasoli,
chichun-charlie-liu,
kcirred,
nwang-ibm and
tharapalanivel
as code owners
bayo-ibm
changed the title
bayo-ibm
changed the title
bayo-ibm
force-pushed
the
eval_1gpu
branch
2 times, most recently
from
8a68e2c to
6cdbd4c
Compare
Signed-off-by: omobayode.fagbohungbe <omobayode.fagbohungbe@ibm.com>
andrea-fasoli
approved these changes
Jul 22, 2025
Collaborator
andrea-fasoli
left a comment
andrea-fasoli
left a comment
There was a problem hiding this comment.
I worked with Bayo to troubleshoot this Granite evaluation issue. The solution looks functional.
Signed-off-by: omobayode.fagbohungbe <omobayode.fagbohungbe@ibm.com>
Signed-off-by: omobayode.fagbohungbe <omobayode.fagbohungbe@ibm.com>
Collaborator
|
pylint issue regarding triton kernel will be fixed by #166 . |
andrea-fasoli
approved these changes
Jul 24, 2025
chichun-charlie-liu
merged commit 78214ae
into
foundation-model-stack:main
10 of 11 checks passed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 of the change
The block-by-block model evaluation does not work with granite-3.x models. This is because the eval_llm_1GPU does not take into consideration the peculiarities associated with granite-3.x-models.
The quant/ptq.py was updated to reflect the current naming conventions for the various blocks in granite-3.x models.
The eval_llm_1GPU function was updated to correctly calculate the logits for granite-3. x models.
Related issues or PRs
#155
How to verify the PR
Was the PR tested
The fix was validated by performing FP8 DQ-SQ using the granite-3.0-8B-Instruct/granite-3.3-8B-Instruct models. The evaluation of the quantized and the unquantized models using eval_llm_1GPU achieved the same results as using the "evaluator.evaluate" method
Checklist for passing CI/CD:
git commit -signoffor equivalenttox -e fixtox -e linttox -e spellchecktox -e unitNote: CI/CD performs unit tests on multiple versions of Python from a fresh install. There may be differences with your local environment and the test environment.