OpInfo: true_divide and minor fix #59154

kshitij12345 · 2021-05-28T17:03:25Z

Reference: #54261

facebook-github-bot · 2021-05-28T17:03:31Z

💊 CI failures summary and remediations

As of commit c08c481 (more details on the Dr. CI page):

✅ None of the CI failures appear to be your fault 💚

1/1 tentatively recognized as flaky ❄️
- Click here to rerun these jobs

❄️ 1 failure tentatively classified as flaky

but reruns have not yet been triggered to confirm:

pytorch_xla_linux_bionic_py3_6_clang9_test (1/1)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun) ❄️

May 29 11:19:16 RuntimeError: tensorflow/compil...OK() (Unknown: Could not start gRPC server vs. OK)

May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch_xla-1.9-py3.6-linux-x86_64.egg/torch_xla/distributed/xla_multiprocessing.py", line 314, in _setup_replication
May 29 11:19:16     device = xm.xla_device()
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch_xla-1.9-py3.6-linux-x86_64.egg/torch_xla/core/xla_model.py", line 232, in xla_device
May 29 11:19:16     devkind=devkind if devkind is not None else None)
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch_xla-1.9-py3.6-linux-x86_64.egg/torch_xla/core/xla_model.py", line 137, in get_xla_supported_devices
May 29 11:19:16     xla_devices = _DEVICES.value
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch_xla-1.9-py3.6-linux-x86_64.egg/torch_xla/utils/utils.py", line 32, in value
May 29 11:19:16     self._value = self._gen_fn()
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch_xla-1.9-py3.6-linux-x86_64.egg/torch_xla/core/xla_model.py", line 19, in <lambda>
May 29 11:19:16     _DEVICES = xu.LazyProperty(lambda: torch_xla._XLAC._xla_get_devices())
May 29 11:19:16 RuntimeError: tensorflow/compiler/xla/xla_client/xrt_local_service.cc:56 : Check failed: tensorflow::NewServer(server_def, &server_) == ::tensorflow::Status::OK() (Unknown: Could not start gRPC server vs. OK)
May 29 11:19:16 Traceback (most recent call last):
May 29 11:19:16   File "/var/lib/jenkins/workspace/xla/test/test_mp_all_to_all.py", line 34, in <module>
May 29 11:19:16     xmp.spawn(_mp_fn, args=())
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch_xla-1.9-py3.6-linux-x86_64.egg/torch_xla/distributed/xla_multiprocessing.py", line 394, in spawn
May 29 11:19:16     start_method=start_method)
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch/multiprocessing/spawn.py", line 188, in start_processes
May 29 11:19:16     while not context.join():
May 29 11:19:16   File "/opt/conda/lib/python3.6/site-packages/torch/multiprocessing/spawn.py", line 144, in join
May 29 11:19:16     exit_code=exitcode
May 29 11:19:16 torch.multiprocessing.spawn.ProcessExitedException: process 1 terminated with exit code 17

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.

Click here to manually regenerate this comment.

kshitij12345 · 2021-05-28T17:04:36Z

torch/testing/_internal/common_methods_invocations.py

-        ('true_divide', (S, S, S), (uniform_scalar(0.1),), 'scalar_broadcast_rhs', (True,)),
-        ('true_divide', (), (uniform_scalar(0.1),), 'scalar_broadcast_lhs', (True,)),
-        ('true_divide', torch.rand(S, S, S) + 1e-1, (3.14,), 'constant', (True,)),
-        ('true_divide', uniform_scalar(1e-1, requires_grad=True), (3.14,), 'scalar_constant', (True,)),


Cases covered by binary_pwise,

pytorch/torch/testing/_internal/common_methods_invocations.py

Lines 727 to 741 in 41054f2

scalar = 3.14 + 3.14j if dtype.is_complex else (3.14 if dtype.is_floating_point else 3)

scalar = 1 if dtype is torch.bool else scalar

tests_list = [

((S, S, S), (S, S, S), False),

((S, S, S), (S, S), False),

((), (), False),

((S, S, S), (), False),

((S, S, S), scalar, False),

((), scalar, False)

]

tests_with_lhs_broadcasting = [

((S, S), (S, S, S), True),

((), (S, S, S), True),

((S, 1, S), (M, S), True),

]

mruberry · 2021-05-29T04:49:05Z

Another NNC issue:

======================================================================
ERROR [0.366s]: test_unsupported_true_divide_cpu_float32 (__main__.TestNNCOpInfoCPU)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 297, in instantiated_test
    raise rte
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 292, in instantiated_test
    result = test_fn(self, *args)
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 266, in test_wrapper
    return test(*args, **kwargs)
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 729, in only_fn
    return fn(slf, device, *args, **kwargs)
  File "test_jit_fuser_te.py", line 2001, in test_unsupported
    raise RuntimeError("Expected test to fail. If it now works, move op into works_list")
RuntimeError: Expected test to fail. If it now works, move op into works_list

cc @Chillee

Also we should rename that test to be more specific, like "test_nnc_unsupported"

mruberry · 2021-05-29T04:50:14Z

Another NNC issue:

======================================================================
ERROR [0.366s]: test_unsupported_true_divide_cpu_float32 (__main__.TestNNCOpInfoCPU)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 297, in instantiated_test
    raise rte
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 292, in instantiated_test
    result = test_fn(self, *args)
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 266, in test_wrapper
    return test(*args, **kwargs)
  File "/opt/conda/lib/python3.6/site-packages/torch/testing/_internal/common_device_type.py", line 729, in only_fn
    return fn(slf, device, *args, **kwargs)
  File "test_jit_fuser_te.py", line 2001, in test_unsupported
    raise RuntimeError("Expected test to fail. If it now works, move op into works_list")
RuntimeError: Expected test to fail. If it now works, move op into works_list

cc @Chillee

Also we should rename that test to be more specific, like "test_nnc_unsupported"

And can we move it into test_ops.py? That way people only have to run one test file when adding an OpInfo. The jit support tests are also in test_ops.py, for example.

torch/testing/_internal/common_methods_invocations.py

mruberry · 2021-05-29T04:53:16Z

torch/testing/_internal/common_utils.py

-            real = torch.tensor(torch.finfo(float_dtype).eps, device=device, dtype=dtype)
-            imag = torch.tensor(torch.finfo(float_dtype).eps, device=device, dtype=dtype)
-            replace_with = torch.complex(real, imag)
+            float_eps = torch.tensor(torch.finfo(float_dtype).eps, device=device, dtype=float_dtype)


mruberry

Nice work @kshitij12345. This needs a fix for the NNC test and a small comment update.

I also left a couple notes for @Chillee to move that test into test_ops.py (if possible), improve its naming, and (these notes are elsewhere) add metadata to OpInfos for whether they support NNC or not. That should make it easier to work with that test. But we should do any of those things in this PR.

Is div next?

Chillee · 2021-05-29T05:03:18Z

I also left a couple notes for @Chillee to move that test into test_ops.py (if possible), improve its naming, and (these notes are elsewhere) add metadata to OpInfos for whether they support NNC or not. That should make it easier to work with that test. But we should do any of those things in this PR.

These changes sound fine to me - the main consideration is just maximizing the ease of adding new OpInfos + maintainibility of these OpInfos.

I'd be glad to put up a PR for it, although I'm PTO the next 2 weeks, so it's possible I won't get to it (unless I do it now... maybe). Feel free to move the code to where you'd like.

krshrimali

LGTM, @kshitij12345!

facebook-github-bot · 2021-05-30T08:01:09Z

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

mruberry · 2021-05-30T08:04:46Z

I also left a couple notes for @Chillee to move that test into test_ops.py (if possible), improve its naming, and (these notes are elsewhere) add metadata to OpInfos for whether they support NNC or not. That should make it easier to work with that test. But we should do any of those things in this PR.

These changes sound fine to me - the main consideration is just maximizing the ease of adding new OpInfos + maintainibility of these OpInfos.

I'd be glad to put up a PR for it, although I'm PTO the next 2 weeks, so it's possible I won't get to it (unless I do it now... maybe). Feel free to move the code to where you'd like.

Enjoy your PTO! I filed #59185 (comment) to track this

facebook-github-bot · 2021-05-31T01:36:47Z

@mruberry merged this pull request in ea465f7.

Summary: Reference: #54261 Depends on: #59154 Pull Request resolved: #59173 Reviewed By: ngimel Differential Revision: D28785178 Pulled By: mruberry fbshipit-source-id: 902310f2d77e499a2355a23b2d5a8c0b21b8c5bb

Summary: Reference: pytorch#54261 Pull Request resolved: pytorch#59154 Reviewed By: ngimel Differential Revision: D28780115 Pulled By: mruberry fbshipit-source-id: 91e254698597fa0c7d4df6053ec017a85e180304

Summary: Reference: pytorch#54261 Depends on: pytorch#59154 Pull Request resolved: pytorch#59173 Reviewed By: ngimel Differential Revision: D28785178 Pulled By: mruberry fbshipit-source-id: 902310f2d77e499a2355a23b2d5a8c0b21b8c5bb

OpInfo: true_divide and minor fix

0a2ce9e

facebook-github-bot added the cla signed label May 28, 2021

kshitij12345 commented May 28, 2021

View reviewed changes

kshitij12345 requested a review from krshrimali May 28, 2021 17:04

kshitij12345 marked this pull request as ready for review May 28, 2021 17:05

kshitij12345 requested a review from mruberry May 28, 2021 17:05

pytorchbot added the open source label May 28, 2021

mruberry reviewed May 29, 2021

View reviewed changes

torch/testing/_internal/common_methods_invocations.py Show resolved Hide resolved

mruberry reviewed May 29, 2021

View reviewed changes

mruberry approved these changes May 29, 2021

View reviewed changes

krshrimali approved these changes May 29, 2021

View reviewed changes

kshitij12345 added 2 commits May 29, 2021 03:20

add to works_list

2bebf90

add comment

c08c481

kshitij12345 mentioned this pull request May 29, 2021

OpInfo: div - port pending method_tests entry #59173

Closed

mruberry mentioned this pull request May 30, 2021

Add supports_nnc metadata to OpInfos #59185

Open

facebook-github-bot closed this in ea465f7 May 31, 2021

facebook-github-bot added the Merged label May 31, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

OpInfo: true_divide and minor fix #59154

OpInfo: true_divide and minor fix #59154

Uh oh!

kshitij12345 commented May 28, 2021

Uh oh!

facebook-github-bot commented May 28, 2021 •

edited

Loading

Uh oh!

kshitij12345 May 28, 2021

Uh oh!

mruberry commented May 29, 2021

Uh oh!

mruberry commented May 29, 2021

Uh oh!

Uh oh!

mruberry May 29, 2021

Uh oh!

mruberry left a comment

Uh oh!

Chillee commented May 29, 2021

Uh oh!

krshrimali left a comment

Uh oh!

facebook-github-bot commented May 30, 2021

Uh oh!

mruberry commented May 30, 2021

Uh oh!

facebook-github-bot commented May 31, 2021

Uh oh!

Uh oh!

	scalar = 3.14 + 3.14j if dtype.is_complex else (3.14 if dtype.is_floating_point else 3)
	scalar = 1 if dtype is torch.bool else scalar
	tests_list = [
	((S, S, S), (S, S, S), False),
	((S, S, S), (S, S), False),
	((), (), False),
	((S, S, S), (), False),
	((S, S, S), scalar, False),
	((), scalar, False)
	]
	tests_with_lhs_broadcasting = [
	((S, S), (S, S, S), True),
	((), (S, S, S), True),
	((S, 1, S), (M, S), True),
	]

OpInfo: true_divide and minor fix #59154

OpInfo: true_divide and minor fix #59154

Uh oh!

Conversation

kshitij12345 commented May 28, 2021

Uh oh!

facebook-github-bot commented May 28, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CI failures summary and remediations

❄️ 1 failure tentatively classified as flaky

pytorch_xla_linux_bionic_py3_6_clang9_test (1/1)

Uh oh!

kshitij12345 May 28, 2021

Choose a reason for hiding this comment

Uh oh!

mruberry commented May 29, 2021

Uh oh!

mruberry commented May 29, 2021

Uh oh!

Uh oh!

mruberry May 29, 2021

Choose a reason for hiding this comment

Uh oh!

mruberry left a comment

Choose a reason for hiding this comment

Uh oh!

Chillee commented May 29, 2021

Uh oh!

krshrimali left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented May 30, 2021

Uh oh!

mruberry commented May 30, 2021

Uh oh!

facebook-github-bot commented May 31, 2021

Uh oh!

Uh oh!

facebook-github-bot commented May 28, 2021 •

edited

Loading