Skip to content

Avoid T5GemmaModelTest::test_eager_matches_sdpa_inference being flaky #40702

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 4, 2025
Merged

Avoid T5GemmaModelTest::test_eager_matches_sdpa_inference being flaky #40702

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions tests/models/t5gemma/test_modeling_t5gemma.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -202,6 +202,15 @@ def prepare_config_and_inputs(self):
input_ids = torch.where(input_ids == self.bos_token_id, 42, input_ids)
decoder_input_ids = torch.where(decoder_input_ids == self.bos_token_id, 42, decoder_input_ids)

# Avoid leading PAD tokens from inputs.
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we probably better to force this for all model tester classes

# `T5GemmaForTokenClassification` and `T5GemmaForSequenceClassification` specify `use_cache=False` when
# calling `self.model`. For `self.use_attention_mask=False` case below, the model goes through
# `make_default_2d_attention_mask`. When there are some pad tokens at the beginning of a sequence, it can't
# attend to any place, and the computed mask `[-3.4028e+38, -3.4028e+38, -3.4028e+38, -3.4028e+38, -3.4028e+38, -3.4028e+38, -3.4028e+38]`
# causes larger differences in some equivalence tests.
# Let's avoid such leading PAD tokens.
decoder_input_ids[:, 0] = self.pad_token_id + 1

attention_mask = None
decoder_attention_mask = None
if self.use_attention_mask:
Expand Down