Skip to content

Fix TransformerEncoderLayer Full Mask UT Failure on XPU #2336

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

Open
wants to merge 6 commits into
base: main
Choose a base branch
from
Open

Fix TransformerEncoderLayer Full Mask UT Failure on XPU #2336

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
b298747
Update test_nn_xpu.py
yucai-intel Nov 11, 2025
9abf073
format
yucai-intel Nov 11, 2025
b6d1398
Update test_nn_xpu.py
yucai-intel Nov 11, 2025
3897e97
format
yucai-intel Nov 12, 2025
54091fd
Update test_nn_xpu.py
yucai-intel Nov 20, 2025
c6b8b5a
Merge branch 'main' into yucai/fix/trans
yucai-intel Nov 20, 2025
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
11 changes: 10 additions & 1 deletion test/xpu/test_nn_xpu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3132,8 +3132,17 @@ def perm_fn(x):
# will result in nan.
mask = torch.tensor([[1]], device=device) == 1
result = model(encoder_input, src_key_padding_mask=mask)
fast_path_device = result.is_xpu or result.is_cpu
result = result.cpu().detach().numpy()
self.assertTrue(np.isnan(result).all())
# Non Fast Paths
if training or not batch_first or TEST_WITH_CROSSREF or not fast_path_device:
# We changed the semenatic, on the non fast path so that fully masked out rows return
# 0 from attention thus NaNs should no longer be present and the output should be nonzero
# due to skip connections
self.assertTrue(not np.isnan(result).any())
else:
# Fast Paths
self.assertTrue(np.isnan(result).all())

# deterministic input
encoder_input = perm_fn(
Expand Down