Adding support for gemma3 TransformerEvalWrapper #3074
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
Summary:
This is to support AWQ for gemma3
Test Plan:
Before change:
```
File "/data/users/jerryzh/ao/.github/scripts/torchao_model_releases/quantize_gemma3.py", line 59, in <module>
TransformerEvalWrapper(
File "/home/jerryzh/.conda/envs/hmbd/lib/python3.10/site-packages/torchao-0.14.0+gitb47f1a36-py3.10.egg/torchao/_models/_eval.py", line 82, in run_eval
result = evaluate(
File "/home/jerryzh/.conda/envs/hmbd/lib/python3.10/site-packages/lm_eval/utils.py", line 456, in _wrapper
return fn(*args, **kwargs)
File "/home/jerryzh/.conda/envs/hmbd/lib/python3.10/site-packages/lm_eval/evaluator.py", line 585, in evaluate
resps = getattr(lm, reqtype)(cloned_reqs)
File "/home/jerryzh/.conda/envs/hmbd/lib/python3.10/site-packages/lm_eval/api/model.py", line 391, in loglikelihood
return self._loglikelihood_tokens(new_reqs, disable_tqdm=disable_tqdm)
File "/home/jerryzh/.conda/envs/hmbd/lib/python3.10/site-packages/lm_eval/models/huggingface.py", line 1279, in _loglikelihood_tokens
multi_logits = F.log_softmax(
File "/home/jerryzh/.conda/envs/hmbd/lib/python3.10/site-packages/torch/nn/functional.py", line 2245, in log_softmax
ret = input.log_softmax(dim)
AttributeError: 'Gemma3CausalLMOutputWithPast' object has no attribute 'log_softmax'
Running loglikelihood requests: 0%|
```
After change:
quantize_qwen3.py: https://gist.github.com/jerryzh168/85cc75bc1feb4723fddb156582adc6ad
Uploaded checkpoint after the change: https://huggingface.co/jerryzh168/gemma-3-12b-it-AWQ-INT4
Reviewers:
Subscribers:
Tasks:
Tags:
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3074
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit feae967 with merge base b47f1a3 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
jerryzh168
added
the
topic: improvement
metascroy
approved these changes
Sep 26, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
Summary:
This is to support AWQ for gemma3
Test Plan:
Before change:
After change:
quantize_qwen3.py: https://gist.github.com/jerryzh168/85cc75bc1feb4723fddb156582adc6ad Uploaded checkpoint after the change: https://huggingface.co/jerryzh168/gemma-3-12b-it-AWQ-INT4
Reviewers:
Subscribers:
Tasks:
Tags: