Add typing annotations for torch.nn.quantized.dynamic.modules.rnn

[guilhermeleobas]

Fixes #43185

xref: gh-43072

[dr-ci]

💊 CI failures summary and remediations

As of commit c5939ad (more details on the Dr. CI page):

---

• 2/2 failures possibly* introduced in this PR
  • 2/2 non-CircleCI failure(s)

---

Extra GitHub checks: 2 failed

• Failed: Codecov - codecov/project
• Failed: GitHub Actions - flake8-py3

---

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 on the GitHub issue tracker or post in the (internal) Dr. CI Users group.

See how this bot performed.

This comment has been revised 71 times.

[hameerabbasi]

Probably rebase or merge viable/strict.

[guilhermeleobas]

It seems that the failures were introduced in this PR. I can't reproduce because it requires fbgemm:

Command to reproduce test failure:

$ pytest test/test_quantization.py -k 'test_quantized_rnn' -sv -rs

platform linux -- Python 3.8.4, pytest-5.4.3, py-1.9.0, pluggy-0.13.1 -- /home/guilhermel/miniconda3/envs/pytorch-cuda-dev/bin/python
cachedir: .pytest_cache
hypothesis profile 'default' -> database=DirectoryBasedExampleDatabase('/work/guilhermel/git/Quansight/pytorch/.hypothesis/examples')
rootdir: /work/guilhermel/git/Quansight/pytorch
plugins: hypothesis-5.20.3
collected 323 items / 321 deselected / 2 selected

test/test_quantization.py::TestPostTrainingDynamic::test_quantized_rnn SKIPPED
test/test_quantization.py::TestPostTrainingDynamic::test_quantized_rnn_cell SKIPPED

======================================================================================================== short test summary info ========================================================================================================
SKIPPED [1] test/quantization/test_quantize.py:737: Quantized operations require FBGEMM. FBGEMM is only optimized for CPUs with instruction set support AVX2 or newer.
SKIPPED [1] test/quantization/test_quantize.py:778: Quantized operations require FBGEMM. FBGEMM is only optimized for CPUs with instruction set support AVX2 or newer.

[hameerabbasi]

Perhaps merge viable/strict?

[guilhermeleobas]

It seems that the failures were introduced in this PR. I can't reproduce because it requires fbgemm:

Command to reproduce test failure:

$ pytest test/test_quantization.py -k 'test_quantized_rnn' -sv -rs

platform linux -- Python 3.8.4, pytest-5.4.3, py-1.9.0, pluggy-0.13.1 -- /home/guilhermel/miniconda3/envs/pytorch-cuda-dev/bin/python
cachedir: .pytest_cache
hypothesis profile 'default' -> database=DirectoryBasedExampleDatabase('/work/guilhermel/git/Quansight/pytorch/.hypothesis/examples')
rootdir: /work/guilhermel/git/Quansight/pytorch
plugins: hypothesis-5.20.3
collected 323 items / 321 deselected / 2 selected

test/test_quantization.py::TestPostTrainingDynamic::test_quantized_rnn SKIPPED
test/test_quantization.py::TestPostTrainingDynamic::test_quantized_rnn_cell SKIPPED

======================================================================================================== short test summary info ========================================================================================================
SKIPPED [1] test/quantization/test_quantize.py:737: Quantized operations require FBGEMM. FBGEMM is only optimized for CPUs with instruction set support AVX2 or newer.
SKIPPED [1] test/quantization/test_quantize.py:778: Quantized operations require FBGEMM. FBGEMM is only optimized for CPUs with instruction set support AVX2 or newer.

I don't think this test are related to the changes introduced in this pull request. I can reproduce the same error on master and viable/strict.

The test fails in the following CI machines

• pytorch_linux_bionic_py3_8_gcc9_test
• pytorch_linux_bionic_py3_6_clang9_test
• pytorch_linux_xenial_py3_clang5_asan_test2
• pytorch_linux_xenial_py3_6_gcc5_4_test
• pytorch_windows_vs2019_py36_cuda10.1_test2
• pytorch_linux_xenial_py3_6_gcc5_4_ge_config_simple_test
• pytorch_linux_xenial_cuda10_2_cudnn7_py3_gcc7_test

[rgommers]

I don't think this test are related to the changes introduced in this pull request. I can reproduce the same error on master and viable/strict.

You mean in CI on this PR? https://ezyang.github.io/pytorch-ci-hud/build/pytorch-master looks very green right now. It's likely caused by this PR.

The 11 JIT failures in pytorch_linux_bionic_py3_6_clang9_test look unrelated, but would be nice to be sure.

[rgommers]

You mean in CI on this PR? https://ezyang.github.io/pytorch-ci-hud/build/pytorch-master looks very green right now. It's likely caused by this PR.

I rebuilt with fbgemm support and tested this branch - the two hypothesis tests pass for me locally.

[rgommers]

@guilhermeleobas other PRs don't have these failures, so some more digging is needed. For starters, can you squash all your commits and rebase on current master?

[guilhermeleobas]

The error was related to pytorch not parsing # type: ignore comments correctly. PR #43446 seems to fix that. I will rebase and run CI again.

[codecov]

Codecov Report

Merging #43186 into master will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master   #43186      +/-   ##
==========================================
- Coverage   69.34%   69.34%   -0.01%     
==========================================
  Files         378      378              
  Lines       46698    46697       -1     
==========================================
- Hits        32381    32380       -1     
  Misses      14317    14317              

Impacted Files	Coverage Δ
torch/nn/quantized/dynamic/modules/rnn.py	89.02% <100.00%> (-0.03%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 26161e8...c5939ad. Read the comment docs.

[rgommers]

LGTM now, thanks @guilhermeleobas

[rgommers]

The two CI failures are unrelated.

[guilhermeleobas]

Thanks for the review, @rgommers and @hameerabbasi

[facebook-github-bot]

@malfet has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[facebook-github-bot]

@malfet merged this pull request in 63a0bb0.