Add type annotations to torch.onnx.* modules #45258

guilhermeleobas · 2020-09-24T03:36:17Z

Still need to resolve a few mypy issues before a review. In special, there is an error which I don't know how to solve, see:

torch/onnx/utils.py:437: error: Name 'is_originally_training' is not defined  [name-defined]
        if training is None or training == TrainingMode.EVAL or (training == TrainingMode.PRESERVE and not is_originally_training):

is_originally_training is used but never defined/imported on torch/onnx/utils.py,

dr-ci · 2020-09-24T03:47:10Z

💊 CI failures summary and remediations

As of commit 1b52305 (more details on the Dr. CI page):

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

Extra GitHub checks: 1 failed

Failed: Codecov - codecov/patch

codecov.io: 1 failed

Failed: codecov/patch

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 63 times.

guilhermeleobas · 2020-09-24T03:52:05Z

torch/onnx/utils.py

@@ -76,7 +77,7 @@ def export(model, args, f, export_params=True, verbose=False, training=None,
    if aten or export_raw_ir:
        assert operator_export_type is None
        assert aten ^ export_raw_ir
-        operator_export_type = OperatorExportTypes.ATEN if aten else OperatorExportTypes.RAW
+        operator_export_type = OperatorExportTypes.ONNX_ATEN if aten else OperatorExportTypes.RAW


There is no mention to OperatorExportTypes.ATEN in the source code. So, I think this value was supposed to be OperatorExportTypes.ONNX_ATEN.

guilhermeleobas · 2020-09-24T03:52:52Z

torch/onnx/utils.py

@@ -466,7 +468,7 @@ def export_to_pretty_string(model, args, f, export_params=True, verbose=False, t
    if aten or export_raw_ir:
        assert operator_export_type is None
        assert aten ^ export_raw_ir
-        operator_export_type = OperatorExportTypes.ATEN if aten else OperatorExportTypes.RAW
+        operator_export_type = OperatorExportTypes.ONNX_ATEN if aten else OperatorExportTypes.RAW


guilhermeleobas · 2020-09-29T19:25:57Z

torch/onnx/utils.py

@@ -432,7 +434,7 @@ def _model_to_graph(model, args, verbose=False,
    param_names = input_and_param_names[len(input_and_param_names) - len(params):]
    params_dict = dict(zip(param_names, params))

-    if training is None or training == TrainingMode.EVAL or (training == TrainingMode.PRESERVE and not is_originally_training):


is_originally_training is not defined anywhere. So, I guess this part of the conditional isn't executed.

rgommers

LGTM, thanks @guilhermeleobas

rgommers · 2020-10-02T22:32:08Z

This has a merge conflict, can you rebase @guilhermeleobas?

guilhermeleobas · 2020-10-07T15:19:34Z

Failures are real

torch/onnx/utils.py:203: error: Module has no attribute "_jit_pass_onnx_set_dynamic_input_shape"  [attr-defined]
torch/onnx/utils.py:225: error: Module has no attribute "_jit_pass_onnx_graph_shape_type_inference"; maybe "_jit_pass_onnx_node_shape_type_inference"?  [attr-defined]
torch/onnx/utils.py:429: error: Module has no attribute "_jit_pass_onnx_assign_output_shape"  [attr-defined]
torch/onnx/utils.py:435: error: Module has no attribute "_jit_pass_onnx_assign_output_shape"  [attr-defined]
torch/onnx/symbolic_opset11.py:806: error: Module has no attribute "_jit_pass_fixup_onnx_loop_node_inputs"; maybe "_jit_pass_fixup_onnx_controlflow_node"?  [attr-defined]

Edit: Fixed! CI should be all green now

rgommers · 2020-10-16T21:48:48Z

@guilhermeleobas there's still CI failures here that are real:

Traceback (most recent call last):
  File "test_type_hints.py", line 217, in test_run_mypy
    self.fail(f"mypy failed: {stdout} {stderr}")
AssertionError: mypy failed: torch/_C/__init__.pyi:375: error: Name 'Value' already defined on line 368  [no-redef]
torch/_C/__init__.pyi:379: error: Name 'Block' already defined on line 371  [no-redef]
torch/_C/__init__.pyi:383: error: Name 'Node' already defined on line 365  [no-redef]
Found 3 errors in 1 file (checked 1099 source files)

Can you update the PR?

guilhermeleobas · 2020-10-21T21:32:57Z

@rgommers those issues should now be fixed.

The current mypy failures are not related to any changes introduced in this PR and there are fixed (ignored) in a different PR

~/git/pytorch:torch-onnx (pytorch-cuda-dev) guilhermel@pytorch-dev $ mypy
torch/distributions/kl.py:107: error: Value of type variable "_LT" of "min" cannot be "_Match"  [type-var]
torch/distributions/kl.py:108: error: Value of type variable "_LT" of "min" cannot be "_Match"  [type-var]
torch/utils/cpp_extension.py:1560: error: Argument 2 to "load_module" has incompatible type "IO[Any]"; expected "Optional[_FileLike]"  [arg-type]
torch/testing/_internal/common_utils.py:1040: error: Value of type variable "_LT" of "max" cannot be "Optional[float]"  [type-var]

rgommers · 2020-10-22T11:32:30Z

The current mypy failures are not related to any changes introduced in this PR and there are fixed (ignored) in a different PR

Yes. Would be nice to get those fixes landed first and then re-run the CI on this PR, because the onnx_ort_test1/2 jobs would be good to see green.

guilhermeleobas · 2020-10-23T20:31:37Z

Failure is not related to this PR:

...
Oct 22 20:00:18 + cleanup
Oct 22 20:00:18 + retcode=0
Oct 22 20:00:18 + set +x
Oct 22 20:00:18 Error response from daemon: No such exec instance: 9d9da493a69d31cab48a6c46379347093d2a2fcf77e3a9a486be69535b1dac3d

Exited with code exit status 1

rgommers · 2020-10-23T21:00:49Z

LGTM now, thanks @guilhermeleobas.

@malfet would you be able to land this PR?

facebook-github-bot

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

facebook-github-bot · 2020-12-02T05:25:09Z

@ezyang merged this pull request in 40a2dd7.

mruberry · 2020-12-02T10:21:20Z

Unlanding. This appears to have broken multiple builds (see the Github CI, too). Sample snippet:

test_run_mypy - TestTypeHints
test_type_hints.py
Traceback (most recent call last):
  File "test_type_hints.py", line 217, in test_run_mypy
    self.fail(f"mypy failed: {stdout} {stderr}")
AssertionError: mypy failed: torch/_C/__init__.pyi:281: error: Name '_jit_pass_constant_propagation' already defined on line 206  [no-redef]
torch/onnx/symbolic_helper.py:368: error: Name 'unbind' already defined (possibly by an import)  [no-redef]
Found 2 errors in 2 files (checked 1157 source files)

guilhermeleobas · 2020-12-02T15:02:56Z

@rgommers done!

codecov · 2020-12-02T18:24:43Z

Codecov Report

Merging #45258 (1b52305) into master (9c6979a) will increase coverage by 0.00%.
The diff coverage is 72.41%.

@@           Coverage Diff           @@
##           master   #45258   +/-   ##
=======================================
  Coverage   80.83%   80.83%           
=======================================
  Files        1859     1859           
  Lines      200633   200639    +6     
=======================================
+ Hits       162177   162182    +5     
- Misses      38456    38457    +1

ezyang · 2020-12-02T19:41:38Z

@guilhermeleobas could you please reopen a new PR? I cna't import this one anymore

guilhermeleobas · 2020-12-02T22:04:09Z

@ezyang, sure!

Summary: Fixes #45215 This is a follow up PR of #45258 Pull Request resolved: #48782 Reviewed By: heitorschueroff Differential Revision: D25304229 Pulled By: ezyang fbshipit-source-id: b01b21ddbf86f908ca08173e68b81fb25851bc81

Summary: Fixes #45215 This is a follow up PR of #45258 and #48782 Pull Request resolved: #48980 Reviewed By: zhangguanheng66 Differential Revision: D25399823 Pulled By: ezyang fbshipit-source-id: 798055f4abbbffecdfab0325884193c81addecec

guilhermeleobas added the module: typing Related to mypy type annotations label Sep 24, 2020

guilhermeleobas self-assigned this Sep 24, 2020

pytorchbot added the open source label Sep 24, 2020

guilhermeleobas commented Sep 24, 2020

View reviewed changes

guilhermeleobas commented Sep 29, 2020

View reviewed changes

guilhermeleobas marked this pull request as ready for review September 29, 2020 19:27

zhangguanheng66 added module: onnx Related to torch.onnx triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module labels Sep 29, 2020

zhangguanheng66 requested a review from houseroad September 29, 2020 20:15

rgommers self-requested a review October 2, 2020 17:41

rgommers approved these changes Oct 2, 2020

View reviewed changes

malfet approved these changes Oct 16, 2020

View reviewed changes

facebook-github-bot added the cla signed label Oct 31, 2020

guilhermeleobas requested a review from malfet November 30, 2020 16:57

rgommers requested a review from ezyang December 1, 2020 15:34

facebook-github-bot reviewed Dec 1, 2020

View reviewed changes

facebook-github-bot closed this in 40a2dd7 Dec 2, 2020

facebook-github-bot added the Merged label Dec 2, 2020

mruberry reopened this Dec 2, 2020

guilhermeleobas removed request for soumith, malfet, albanD, rohan-varma, fmassa, pritamdamania87, mrshenli, houseroad, glaringlee, mingzhe09088 and zhaojuanmao December 2, 2020 14:42

guilhermeleobas and others added 8 commits December 2, 2020 14:50

Add type annotations to torch.onnx.* modules

22c31d2

fix mypy errors

26ef9f4

fix flake8 error

41911bd

add missing type annotations

c9653e2

update __init__.pyi.in

6d28a76

Replace Set[Any] by Set[int]

7c81308

remove Any import

309facb

remove duplicate and add type: ignore

1b52305

guilhermeleobas closed this Dec 2, 2020

guilhermeleobas mentioned this pull request Dec 3, 2020

Add type annotations to torch.onnx.* modules #48782

Closed

guilhermeleobas mentioned this pull request Dec 8, 2020

Torch onnx #48980

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add type annotations to torch.onnx.* modules #45258

Add type annotations to torch.onnx.* modules #45258

guilhermeleobas commented Sep 24, 2020

dr-ci bot commented Sep 24, 2020 •

edited

guilhermeleobas Sep 24, 2020

guilhermeleobas Sep 24, 2020

guilhermeleobas Sep 29, 2020 •

edited

rgommers left a comment

rgommers commented Oct 2, 2020

guilhermeleobas commented Oct 7, 2020 •

edited

rgommers commented Oct 16, 2020

guilhermeleobas commented Oct 21, 2020

rgommers commented Oct 22, 2020

guilhermeleobas commented Oct 23, 2020 •

edited

rgommers commented Oct 23, 2020

facebook-github-bot left a comment

facebook-github-bot commented Dec 2, 2020

mruberry commented Dec 2, 2020

guilhermeleobas commented Dec 2, 2020

codecov bot commented Dec 2, 2020

ezyang commented Dec 2, 2020

guilhermeleobas commented Dec 2, 2020

Add type annotations to torch.onnx.* modules #45258

Add type annotations to torch.onnx.* modules #45258

Conversation

guilhermeleobas commented Sep 24, 2020

dr-ci bot commented Sep 24, 2020 • edited

💊 CI failures summary and remediations

Extra GitHub checks: 1 failed

codecov.io: 1 failed

guilhermeleobas Sep 24, 2020

Choose a reason for hiding this comment

guilhermeleobas Sep 24, 2020

Choose a reason for hiding this comment

guilhermeleobas Sep 29, 2020 • edited

Choose a reason for hiding this comment

rgommers left a comment

Choose a reason for hiding this comment

rgommers commented Oct 2, 2020

guilhermeleobas commented Oct 7, 2020 • edited

rgommers commented Oct 16, 2020

guilhermeleobas commented Oct 21, 2020

rgommers commented Oct 22, 2020

guilhermeleobas commented Oct 23, 2020 • edited

rgommers commented Oct 23, 2020

facebook-github-bot left a comment

Choose a reason for hiding this comment

facebook-github-bot commented Dec 2, 2020

mruberry commented Dec 2, 2020

guilhermeleobas commented Dec 2, 2020

codecov bot commented Dec 2, 2020

Codecov Report

ezyang commented Dec 2, 2020

guilhermeleobas commented Dec 2, 2020

dr-ci bot commented Sep 24, 2020 •

edited

guilhermeleobas Sep 29, 2020 •

edited

guilhermeleobas commented Oct 7, 2020 •

edited

guilhermeleobas commented Oct 23, 2020 •

edited