Enforce both input tensor shapes of CosineEmbeddingLoss to be equal.

[tringwald]

…Added a test to prevent regressions.

Fixes #112732.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112782

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ You can merge normally! (3 Unrelated Failures)

As of commit e4de008 with merge base 68dead4 ():

UNSTABLE - The following jobs failed but were likely due to flakiness present on trunk and has been marked as unstable:

• trunk / macos-12-py3-arm64 / test (default, 1, 3, macos-m1-12, unstable) (gh)
• trunk / macos-12-py3-arm64 / test (default, 2, 3, macos-m1-12, unstable) (gh)
• trunk / macos-12-py3-arm64 / test (default, 3, 3, macos-m1-12, unstable) (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[lezcano]

The previous code did allow broadcasting. Perhaps we want to allow broadcasting as well?

[tringwald]

The previous code did allow broadcasting. Perhaps we want to allow broadcasting as well?

I've changed the checking logic a bit. Looks like we just need to make sure the second tensor has the same number of dimensions as the first one. Things like this never worked anyway:

>>> torch.__version__
'2.1.0'
>>> torch.nn.functional.cosine_embedding_loss(torch.ones(2, 10), torch.ones(10), target=torch.ones(1))
[...]
IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)
>>> torch.nn.functional.cosine_embedding_loss(torch.ones(10), torch.ones(2, 10), target=torch.ones(1))
RuntimeError: 1D target tensor expects 2D input tensors, but found inputs with sizes [10] and [2, 10].

Should we mention broadcasting semantics in the docs too? Right now it says Input2: (N,D) or (D), same shape as Input1.

[lezcano]

LGTM. Updating the docs to shortly mention broadcasting would also be good.

[tringwald]

@pytorchbot merge

[pytorchmergebot]

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here