OpInfo: log_softmax #59336
OpInfo: log_softmax #59336
Conversation
💊 CI failures summary and remediationsAs of commit 714e727 (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions to the (internal) Dr. CI Users group. |
jbschlosser
added
the
triaged
| 'log_softmax', | ||
| variant_test_name='dtype', | ||
| supports_out=False, | ||
| dtypes=all_types_and_complex_and(torch.bool, torch.float16, torch.bfloat16), |
There was a problem hiding this comment.
Not sure, as to how complex is also supported with dtype=torch.float64.
There was a problem hiding this comment.
Looks like this part of code facilitates that,
pytorch/aten/src/ATen/native/SoftMax.cpp
Lines 273 to 274 in c3bf42e
>>> t = torch.randn(3,3, dtype=torch.complex128)
>>> torch.log_softmax(t, 0, dtype=torch.float64)
<stdin>:1: UserWarning: Casting complex values to real discards the imaginary part (Triggered internally at ../aten/src/ATen/native/Copy.cpp:240.)It does raise a warning. But I think it should just error out.
There was a problem hiding this comment.
This is really interesting -- thanks for taking a closer look. I don't think we should take any action for this behavior in this PR, however.
fyi @anjali411
|
@mruberry has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Summary: Reference: pytorch#54261 Pull Request resolved: pytorch#59336 Reviewed By: agolynski Differential Revision: D28899052 Pulled By: mruberry fbshipit-source-id: 60a9a4ffbca5a0f2c899d4d83500dcab4555ffb0
5 participants
Reference: #54261