Update torch/nn/modules/conv.py to use Literal for support padding modes
#152280
Comments
|
can I work on this one? @Skylion007 |
|
Sure |
|
Hi! I had asked to work on this issue earlier (see above), and received approval. Would it be possible to coordinate next time to avoid duplication? |
|
Hi @Skylion007, I understand someone else submitted a PR earlier, but since my implementation includes test verification and follows the suggested corrections, I hope it can still be considered. Appreciate your time reviewing this, and happy to make any changes needed. Thanks! |
|
Hi @Skylion007 , I'd be happy to take a look at the current mypy type error and help resolve the linting failure related to Literal[...] usage in conv.py. Let me know if it's alright for me to jump in and contribute a patch on top of the current implementation. Thanks! |
* Defines PaddingMode alias
* Annotates Conv{1,2,3}d constructors
* Fixes pytorch#152280
|
"HI i'd like to work on this.Can I take it? |
|
The following files also have similar type annotation issues , torch/ao/nn/qat/modules/conv.py and torch/ao/nn/quantized/modules/conv.py . The QAT version has different padding mode support. The quantized version has a restricted set of supported padding modes (_SUPPORTED_PADDING = {"zeros", "reflect"}) should we file issues to fix those too ? @Skylion007 |
🚀 The feature, motivation and pitch
It would be great to update
torch/nn/modules/conv.pyto use typing.Literal instead of juststrto denote with padding modes are actually supported by various operations.for example instead of
padding_mode : strdo
padding_mode: Literal["valid", "same"]etc to the type checker can catch bugs for the code is actually ran.Alternatives
No response
Additional context
No response
cc @ezyang @malfet @xuzhao9 @gramster
The text was updated successfully, but these errors were encountered: