-
Notifications
You must be signed in to change notification settings - Fork 25.5k
[JIT] add unsupported section #31329
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.
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
Conversation
💊 CircleCI build failures summary and remediationsAs of commit feba4b4:
Detailed failure analysisOne may explore the probable reasons each build failed interactively on the Dr. CI website. 🕵️ 9 new failures recognized by patternsThe following build failures don't appear to be due to upstream breakage:
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Seems like a good start but it's missing a lot of stuff from the quip (which itself is missing some other stuff like the unsupported parts of torch.nn.init
)
docs/source/jit_unsupported.rst
Outdated
|
||
TorchScript supports most methods defined on `torch` and `torch.Tensor`, but we do not have full coverage. | ||
Here are specific known ops and categories of ops which have diverging behavior between | ||
Python and TorchScript. Deprecated ops are not listed below. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe add a note that says if you encounter something not listed here that's not supported you should file a GitHub issue
we suggest using :meth:`torch.jit.trace`. | ||
|
||
* :class:`torch.nn.RNN` | ||
* :class:`torch.nn.AdaptiveLogSoftmaxWithLoss` |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is missing torch.nn.modules.normalization.CrossMapLRN2d
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
CrossMapLRN2d
isn't documented
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
CrossMapLRN2d
isn't in our documentation
docs/source/jit_unsupported.rst
Outdated
|
||
TorchScript supports most methods defined on `torch` and `torch.Tensor`, but we do not have full coverage. | ||
Here are specific known ops and categories of ops which have diverging behavior between | ||
Python and TorchScript. Deprecated ops are not listed below. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why not list deprecated ops? We are detecting them below, we could also tag them with something like * resize (this method is deprecated and won't be supported)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sure, we can do that
return x.{op}() | ||
''') | ||
|
||
deprecated_apis = set(["volatile", "resize", "reinforce", "new", "name", "map2_", "has_names", "grad_fn", "resize_as"]) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also where is this list coming from?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
These are the undocumented apis & properties. I could try to expose the list here: https://github.com/pytorch/pytorch/blob/master/test/test_docs_coverage.py#L83
What stuff is missing from the quip exactly? Also |
FYI only the non-inplace functions in torch.nn.init are are deprecated |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@eellison has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Summary: Add a section for unsupported ops, and modules. Automatically generate the properties and attributes that aren't bound, and for ops that have semantic mismatches set up tests so the docs stay up to date. Pull Request resolved: pytorch#31329 Differential Revision: D19164472 Pulled By: eellison fbshipit-source-id: 46290bb8a64d9de928cfb1eda5ff4558c3799c88
Summary: It was added in pytorch#31329 and removed in a bad merge in pytorch#31138 Pull Request resolved: pytorch#31486 Differential Revision: D19181967 Pulled By: eellison fbshipit-source-id: 7e4b4a9b2042c30ec18f7f737bc4a9a56fac7d92
Add a section for unsupported ops, and modules. Automatically generate the properties and attributes that aren't bound, and for ops that have semantic mismatches set up tests so the docs stay up to date.