Michael Carilli [Fri, 3 Sep 2021 20:21:23 +0000 (13:21 -0700)]
[CUDA graphs] Error if attempting to capture uncapturable nccl (#64440)
Summary:
NCCL < 2.9.6 is not capturable. Attempting to capture it can cause nasty behavior (for example, ive seen capture succeed, but replay silently hang). Pytorch should preempt this with a friendlier error.
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse agolynski SciPioneer H-Huang mrzzd cbalioglu gcramer23
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64440
Reviewed By: mruberry
Differential Revision:
D30733884
Pulled By: ngimel
fbshipit-source-id:
5f2df3cf5cc0e5e68f49bf22a80d9f58064dc7ec
Nikita Shulga [Fri, 3 Sep 2021 17:21:01 +0000 (10:21 -0700)]
Fix logical typo in _compare_trilu_indices (#64468)
Summary:
I'm pretty sure that repeating the same call twice is pretty meaningless and intend was to call `tril`/`tril_indices` in first case and `triu`/`triu_indices` in another
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64468
Reviewed By: mruberry
Differential Revision:
D30744978
Pulled By: malfet
fbshipit-source-id:
7cd36789a7ebf1cc263fb2d875e479c05e7588a4
Ansley Ussery [Fri, 3 Sep 2021 13:10:37 +0000 (06:10 -0700)]
Support Union in TorchScript (#64234)
Summary:
This PR is created to replace https://github.com/pytorch/pytorch/pull/53180 PR stack, which has all the review discussions. Reason for needing a replacement is due to a messy Sandcastle issue.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64234
Reviewed By: gmagogsfm
Differential Revision:
D30656444
Pulled By: ansley
fbshipit-source-id:
77536c8bcc88162e2c72636026ca3c16891d669a
Kefei Lu [Fri, 3 Sep 2021 06:03:02 +0000 (23:03 -0700)]
Add fx2trt pass for removing duplicate output args (#64461)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64461
Fx2TRT does not support duplicate nodes in the output args tuple.
This pass removes duplicate output args from the target subnets and fixes their uses in the top level module where the subnets are called. This pass must be called after acc split on the top-level net and subsequent calls to the acc trace on the subnets.
This pass will change both the subnets and top level module.
Test Plan:
Run:
```
buck run mode/opt -c python.package_style=inplace //caffe2/torch/fb/fx2trt/tests/passes/:test_remove_duplicate_output_args
```
Reviewed By: yinghai
Differential Revision:
D30740499
fbshipit-source-id:
98459f7677980b21c7bffda918158001285572db
Elias Ellison [Fri, 3 Sep 2021 05:16:22 +0000 (22:16 -0700)]
Add fusion enabled apis (#64429)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64429
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision:
D30740897
Pulled By: eellison
fbshipit-source-id:
446aa63b5d763f1cfffea62547db7294368e3438
Elias Ellison [Fri, 3 Sep 2021 05:16:22 +0000 (22:16 -0700)]
update optimize_for_inference docs (#64428)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64428
Test Plan: Imported from OSS
Reviewed By: pbelevich
Differential Revision:
D30740898
Pulled By: eellison
fbshipit-source-id:
b94d2c3deb661a6ba048f19e8c1d5e1799667eeb
James Reed [Fri, 3 Sep 2021 04:11:57 +0000 (21:11 -0700)]
[resubmit][FX] Prototype for guarding against mutable operations in tracing (#64467)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64467
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision:
D30744870
Pulled By: jamesr66a
fbshipit-source-id:
fc652f8b17748f90dbeb83fabf3bd5bb57d6ff1a
Mike Ruberry [Fri, 3 Sep 2021 03:51:38 +0000 (20:51 -0700)]
Skips layer norm OpInfo on tbb platform (#64469)
Summary:
The OpInfo tests appear to be discovering a layer norm x tbb issue that requires investigation. Skipping tests on that platform for now to restore CI signal.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64469
Reviewed By: ngimel
Differential Revision:
D30745746
Pulled By: mruberry
fbshipit-source-id:
282484cc00b867fac85b7df61430d64277da6421
Peter Bell [Fri, 3 Sep 2021 00:43:59 +0000 (17:43 -0700)]
THC: Cleanup dead code (#64441)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64441
Test Plan: Imported from OSS
Reviewed By: gchanan
Differential Revision:
D30735342
Pulled By: ngimel
fbshipit-source-id:
84ab36f7aec6b8cd7f1f34c19a58a382c06ad68d
driazati [Fri, 3 Sep 2021 00:09:48 +0000 (17:09 -0700)]
Regenerate generated github workflows (#64465)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64465
These were out of date and causing master failures
Test Plan: Imported from OSS
Reviewed By: zhouzhuojie
Differential Revision:
D30744594
Pulled By: driazati
fbshipit-source-id:
09a21c3c5d9bc83b368d66cabbafd1ba83302dd3
David Riazati [Thu, 2 Sep 2021 23:58:59 +0000 (16:58 -0700)]
Revert
D30732630: [quant] Enable jit tracing on quantizable LSTM
Test Plan: revert-hammer
Differential Revision:
D30732630 (https://github.com/pytorch/pytorch/commit/
116142143cc2d66c7e582d9f96e00862456fd736)
Original commit changeset:
443e351ebb0e
fbshipit-source-id:
49001392f01366f3b1ccc31139f824c80b86cd40
Zafar Takhirov [Thu, 2 Sep 2021 23:58:36 +0000 (16:58 -0700)]
Revert
D30055886: [quant] AO migration of the `quantize.py`
Test Plan: revert-hammer
Differential Revision:
D30055886 (https://github.com/pytorch/pytorch/commit/
44e3ed88c9a1bd9ee6b0168ba5271a2c6b006cc8)
Original commit changeset:
8ef7470f9fa6
fbshipit-source-id:
c5bd3ead43a2d44b9e56872ec5bd7a195bdac725
Jane Xu [Thu, 2 Sep 2021 23:21:52 +0000 (16:21 -0700)]
[POC] .github: Add event name to concurrency (#64402)
Summary:
This would ensure that manually/API triggered workflows would not cancel other triggered workflows. For example, the manually triggered periodic 11.1 linux job cancelled the scheduled one here, which we may not want:
![image](https://user-images.githubusercontent.com/
31798555/
131752175-
1c99d56e-d344-46e1-b8ac-
9c12bba0569a.png).
This would be helpful later as we use more dispatched workflows (e.g., for bisect functionality)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64402
Reviewed By: malfet
Differential Revision:
D30734860
Pulled By: janeyx99
fbshipit-source-id:
220016716094666e9af836fcd716dd529cf23d8a
Garrett Cramer [Thu, 2 Sep 2021 23:11:10 +0000 (16:11 -0700)]
update rpc tensorpipe logic for sparse tensors (#62960)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62960
A bug was filed a few years ago for sending sparse tensor over rpc #30807.
This pr updates rpc/tensorpipe logic for CUDA sparse tensors. During the serialization process, the pickler.cpp implementation breaks down the sparse tensor into two tensors and metadata. torch/csrc/distributed/rpc/tensorpipe_agent.cpp needs to be updated because it does not have logic sparse tensors. It pushes a single device for a sparse tensor. This is wrong because after the sparse tensor has been serialized, there will be two tensors. The second tensor will not have a device. This will cause the second tensor to have the wrong target device. tensorpipe_utils.cpp needs to be updated because deserialization happens after the data is received on the target pipe. This takes the two tensors and metadata sent and rebuilds the sparse tensor. There will be two tpDescriptors but only one tensor after deserialization. The logic is updated to verify the sparse tensor is on the correct device using the first tpDescriptor.
This pr also updates ivalue.cpp and ivalue.h to support more paths for Sparse COO tensors.
I tested these changes by adding sparse tests to rpc_test.py and dist_autograd_test.py.
Test Plan: Imported from OSS
Reviewed By: gchanan
Differential Revision:
D30717285
Pulled By: gcramer23
fbshipit-source-id:
daee9a56764550f56b131f9dd8e74e23113d6714
Eli Uriegas [Thu, 2 Sep 2021 23:06:17 +0000 (16:06 -0700)]
Revert
D30675780: [FX] Prototype for guarding against mutable operations in tracing
Test Plan: revert-hammer
Differential Revision:
D30675780 (https://github.com/pytorch/pytorch/commit/
795387477fe90e03cb598f3077a32222896e65dd)
Original commit changeset:
b2116b51dcc8
fbshipit-source-id:
d4f1173f4989556ea54974f4c2739ef85a705fae
Zafar Takhirov [Thu, 2 Sep 2021 22:56:54 +0000 (15:56 -0700)]
[quant] Enable jit tracing on quantizable LSTM (#64438)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64438
The quantizable LSTM didn't support jit tracing because it had several non taceable paths. We sacrifice some of the user experience to enable the tracing.
The main UX feature removed is a user-friendly message when trying to access the backwards path in a bidirectional LSTM: When the bidirectional flag is `False`, we used to throw a nice error message when the user tried accessing backwards weights. Now the message is default (removed properties).
Test Plan: `buck test mode/dev //caffe2/test:quantization -- test_custom_module_lstm`
Reviewed By: mtl67
Differential Revision:
D30732630
fbshipit-source-id:
443e351ebb0e2b636c86dea9691b9bf42ffe618f
James Reed [Thu, 2 Sep 2021 22:15:24 +0000 (15:15 -0700)]
[FX] Prototype for guarding against mutable operations in tracing (#64295)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64295
Test Plan: Imported from OSS
Reviewed By: zou3519
Differential Revision:
D30675780
Pulled By: jamesr66a
fbshipit-source-id:
b2116b51dcc87357f0c84192c4c336680875e27a
Eli Uriegas [Thu, 2 Sep 2021 21:49:47 +0000 (14:49 -0700)]
.github: Migrate pytorch_linux_bionic_py_3_6_clang9 to GHA (#64218)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64218
Relies on https://github.com/fairinternal/pytorch-gha-infra/pull/11
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
cc ezyang seemethere malfet walterddr lg20987 pytorch/pytorch-dev-infra bdhirsh
Test Plan: Imported from OSS
Reviewed By: malfet, H-Huang, janeyx99
Differential Revision:
D30651516
Pulled By: seemethere
fbshipit-source-id:
e5843dfe84f096f2872d88f2e53e9408ad2fe399
Erjia Guan [Thu, 2 Sep 2021 20:35:05 +0000 (13:35 -0700)]
Switch Shuffler to use iter-local buffer (#64195)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64195
Test Plan: Imported from OSS
Reviewed By: H-Huang
Differential Revision:
D30642947
Pulled By: ejguan
fbshipit-source-id:
d4b52479b4ae37ad693388b9cdb8eed83a136474
Nikita Shulga [Thu, 2 Sep 2021 20:30:51 +0000 (13:30 -0700)]
Disable CircleCI ROCm build (#64434)
Summary:
Per jithunnair-amd suggestion
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64434
Reviewed By: seemethere, janeyx99
Differential Revision:
D30732289
Pulled By: malfet
fbshipit-source-id:
1932d0a7d1e648006f8030c8237b187d0709f688
Kevin Tse [Thu, 2 Sep 2021 20:06:18 +0000 (13:06 -0700)]
[DataPipe] removing filter's inheritance from map (#64404)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64404
This PR remove `filter`'s inheritance from `map`. This allows `filter` to not have a `__len__` function and that behavior is what we would like.
cc VitalyFedyunin ejguan
Test Plan: Imported from OSS
Reviewed By: gchanan
Differential Revision:
D30713120
Pulled By: NivekT
fbshipit-source-id:
4d5d07555297ee2bd4b49842c0d26cdc00638f6c
Kevin Tse [Thu, 2 Sep 2021 20:06:18 +0000 (13:06 -0700)]
[DataPipe] adding/removing __len__ for different DataPipe (#64398)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64398
cc VitalyFedyunin ejguan
Test Plan: Imported from OSS
Reviewed By: ejguan
Differential Revision:
D30710437
Pulled By: NivekT
fbshipit-source-id:
524eda43a2faa0db0c1a662bf9bb4283f0ade83c
Erjia Guan [Thu, 2 Sep 2021 19:25:15 +0000 (12:25 -0700)]
Fix test_ind_worker_queue by setting max_num_worker based on system resource (#63779)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63779
Fixes #63657
Test Plan: Imported from OSS
Reviewed By: gchanan
Differential Revision:
D30494185
Pulled By: ejguan
fbshipit-source-id:
d1bd24299b25d589889604aaf18ad347bdff4df4
Thomas J. Fan [Thu, 2 Sep 2021 19:15:03 +0000 (12:15 -0700)]
ENH Adds test and docs for modules that already support no batch dims (#62729)
Summary:
Towards https://github.com/pytorch/pytorch/issues/60585
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62729
Reviewed By: H-Huang
Differential Revision:
D30669546
Pulled By: jbschlosser
fbshipit-source-id:
c771c98c1fd9d28fa984b72893585c738c736505
Rohan Varma [Thu, 2 Sep 2021 18:37:54 +0000 (11:37 -0700)]
[DDP] Fix logging iterations (#64411)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64411
These are not actually the training iterations, but are offset by how
frequently DDP stats collection actually runs (default being
kDDPRuntimeLoggingSampleRate = 100). So with this change, they are actually
logged to scuba every:
10, 10 * 100, 40 * 100, etc iterations.
Test Plan: CI
Reviewed By: zhaojuanmao
Differential Revision:
D30718274
fbshipit-source-id:
146bd2428753c93363bee37e487f40104fce3c18
Eli Uriegas [Thu, 2 Sep 2021 18:23:38 +0000 (11:23 -0700)]
.github: Move squid vars to common vars (#64436)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64436
Moves the squid variables to our common jinja template so that when we
have to update them they're all in the same place.
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
cc ezyang seemethere malfet lg20987 pytorch/pytorch-dev-infra
Test Plan: Imported from OSS
Reviewed By: malfet, zhouzhuojie
Differential Revision:
D30732776
Pulled By: seemethere
fbshipit-source-id:
22e3757c4eec775baa8abbaac2ba2a0c69c2b2a9
Eli Uriegas [Thu, 2 Sep 2021 18:23:38 +0000 (11:23 -0700)]
.github: Move upload-artifact-s3 to common var (#64435)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64435
Move upload-artifact-s3 to a common variable to be used amongst our
jinja templates, this should make it easier in the future to update
these images
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
cc ezyang seemethere malfet lg20987 pytorch/pytorch-dev-infra
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision:
D30732777
Pulled By: seemethere
fbshipit-source-id:
51cd485f5abae134c3c49dfa878e6303ba8e5f25
Richard Zou [Thu, 2 Sep 2021 18:06:34 +0000 (11:06 -0700)]
nn.functional.linear OpInfo (#61971)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/61971
Test Plan: - wait for tests
Reviewed By: heitorschueroff
Differential Revision:
D30013750
Pulled By: zou3519
fbshipit-source-id:
ca41dbd98176c12e50ad1410a658f4b06fe99a1e
Eli Uriegas [Thu, 2 Sep 2021 17:56:57 +0000 (10:56 -0700)]
Revert
D30468409: Add fx2trt pass for removing duplicate output args
Test Plan: revert-hammer
Differential Revision:
D30468409 (https://github.com/pytorch/pytorch/commit/
6da7552a8eaae6b85e271bf3edac2fa2ae9f1148)
Original commit changeset:
b4d91b76ab5d
fbshipit-source-id:
e138dc425fe55ffe3585ea5fac4db476931bafed
Hui Guo [Thu, 2 Sep 2021 17:40:02 +0000 (10:40 -0700)]
[tensorexpr] Wrap error msgs with buildErrorMessages for internal asserts (#64409)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64409
Test Plan: Imported from OSS
Reviewed By: navahgar
Differential Revision:
D30717786
Pulled By: huiguoo
fbshipit-source-id:
a3b147d339ff4927f14efa24407cd3b63d80001d
Kefei Lu [Thu, 2 Sep 2021 17:38:43 +0000 (10:38 -0700)]
Add fx2trt pass for removing duplicate output args (#64433)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64433
Fx2TRT does not support duplicate nodes in the output args tuple.
This pass removes duplicate output args from the target subnets and fixes their uses in the top level module where the subnets are called. This pass must be called after acc split on the top-level net and subsequent calls to the acc trace on the subnets.
This pass will change both the subnets and top level module.
Test Plan:
Run:
```
buck run mode/opt -c python.package_style=inplace //caffe2/torch/fb/fx2trt/tests/passes/:test_remove_duplicate_output_args
```
Reviewed By:
842974287
Differential Revision:
D30468409
fbshipit-source-id:
b4d91b76ab5d8a5275d68dd48d1327a44c22568e
Jane Xu [Thu, 2 Sep 2021 16:50:56 +0000 (09:50 -0700)]
CI: Enable using labels to control GHA workflows (#64314)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/62852
Sets a global environment variable containing a list of PR labels. For this PR, the PR_LABELS variable looks like:
```
[
"cla signed",
"ciflow/default"
]
```
confirmed in a run: https://github.com/pytorch/pytorch/runs/
3490072161?check_suite_focus=true
This information can be used in other workflow steps to control the logic. For example, if I want to force a build, I can label my PR with "force-build" and do something like the following in my build script:
```
if [[ "${PR_LABELS}" = *force-build* ]]; then
python setup.py install
else
#use cached wheel or something
fi
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64314
Reviewed By: driazati
Differential Revision:
D30714570
Pulled By: janeyx99
fbshipit-source-id:
80b060ee32643ddd22eb7b8ec548579c7ccf6441
Nicolas Hug [Thu, 2 Sep 2021 16:27:44 +0000 (09:27 -0700)]
Fixes and details to torchhub docs (#63783)
Summary:
This PR:
- adds a few details regarding the newly added `skip_validation` parameter https://github.com/pytorch/pytorch/pull/62139
- uses double-backticks instead of single-backticks since this is rst, not mardown.
- adds a few minor doc nits here and there
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63783
Reviewed By: zou3519
Differential Revision:
D30696658
Pulled By: NicolasHug
fbshipit-source-id:
6f01c7eb3cfcd7e17e4c33c09d193054fa18ad36
Thomas J. Fan [Thu, 2 Sep 2021 16:02:35 +0000 (09:02 -0700)]
TST Adds __repr__ and str to module info (#63737)
Summary:
Follow up to https://github.com/pytorch/pytorch/pull/61935
This PR adds `test_repr` to `test_modules`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63737
Reviewed By: gchanan
Differential Revision:
D30729642
Pulled By: jbschlosser
fbshipit-source-id:
c11a28bc0739abd3ed40727389dd28ed4069edad
Zhaoheng Ni [Thu, 2 Sep 2021 15:59:53 +0000 (08:59 -0700)]
Fix torch.istft length mismatch and window runtime error (#63469)
Summary:
The PR fixes two issues:
- See https://github.com/pytorch/pytorch/issues/62747 and https://github.com/pytorch/audio/issues/1409. The length mismatch when the given ``length`` parameter is longer than expected. Add padding logic in consistent with librosa.
- See https://github.com/pytorch/pytorch/issues/62323. The current implementations checks if the min value of window_envelop.abs() is greater than zero. In librosa they normalize the signal on non-zero values by indexing. Like
```
approx_nonzero_indices = ifft_window_sum > util.tiny(ifft_window_sum)
y[approx_nonzero_indices] /= ifft_window_sum[approx_nonzero_indices]
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63469
Reviewed By: fmassa
Differential Revision:
D30695827
Pulled By: nateanl
fbshipit-source-id:
d034e53f0d65b3fd1dbd150c9c5acf3faf25a164
Mike Iovine [Thu, 2 Sep 2021 15:12:48 +0000 (08:12 -0700)]
[Static Runtime] Add sign/abs/lop1p/mul fusion pass (#64209)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64209
Add a new fusion pass that turns transforms the following pattern:
```
graph(%input):
%0 : Tensor = aten::sign(%input)
%1 : Tensor = aten::abs(%input)
%2 : Tensor = aten::log1p(%1)
%res : Tensor = aten::mul(%0, %2)
return (%res)
```
Into a single op:
```
graph(%input):
%res : Tensor = static_runtim::signed_log1p(%input)
return (%res)
```
The intent is to reduce the number of passes over the tensor. However, enabling this pass actually causes a performance regression, probably due to a lack of vectorization in the fused implementation. Because of this issue, this diff **does not** enable this pass.
Followup: navahgar will add an NNC kernel which is faster than the the unfused version and enable this pass. We still need this version as a fallback since the NNC kernel will not support all dtypes.
Test Plan:
`buck test caffe2/benchmarks/static_runtime:static_runtime_cpptest -- SignedLog1p`
Test passed with new graph pass disabled and enabled.
Reviewed By: hlu1
Differential Revision:
D30559929
fbshipit-source-id:
e4e080cb2e6a705cfdde1fc98bee92b723f8132a
CodemodService FBSourceClangFormatLinterBot [Thu, 2 Sep 2021 15:10:37 +0000 (08:10 -0700)]
[AutoAccept][Codemod][FBSourceClangFormatLinter] Daily `arc lint --take CLANGFORMAT`
Reviewed By: zertosh
Differential Revision:
D30710635
fbshipit-source-id:
e8dae05a7e3a19d656067a4f102aab4a3c93ac42
Seth Elliott [Thu, 2 Sep 2021 14:48:47 +0000 (07:48 -0700)]
Fix broken caffe2 test: PlanExecutorTest.BlockingErrorPlan (#64401)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64401
PlanExecutorTest.BlockingErrorPlan uses `ASSERT_DEATH` which internally performs a `fork()`. This can cause problems under certain configurations that use threads. This change updates this test to use the "threadsafe" style for GTest death tests in order to improve its quality in multithreaded environments.
Test Plan:
I confirmed that this change fixes the issue on my devvm with the following command:
```
buck test mode/dev //caffe2/caffe2:caffe2_test_cpu -- PlanExecutorTest.BlockingErrorPlan
```
Reviewed By: praihan
Differential Revision:
D30709447
fbshipit-source-id:
12ffd9ad0371e2e5b43a9873c80568e5ab02d246
Michael Dagitses [Thu, 2 Sep 2021 13:49:09 +0000 (06:49 -0700)]
simplify op name determination into a single forward pass (#64261)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64261
Note that this does not preserve byte-for-byte compatibility with
existing names.
Test Plan:
* Rely on CI to catch gross errors.
* Merge after release cut to catch subtle issues.
Reviewed By: albanD
Differential Revision:
D30700647
Pulled By: dagitses
fbshipit-source-id:
7b02f34b8fae3041240cc78fbc6bcae498c3acd4
Vasiliy Kuznetsov [Thu, 2 Sep 2021 13:12:07 +0000 (06:12 -0700)]
fix copy.deepcopy on LinearPackedParams (#64367)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64367
This is the same thing as https://github.com/pytorch/pytorch/pull/56154
but for quantized linear. It fixes the behavior of `copy.deepcopy` on
these modules. Before this PR, copied instances of `LinearPackedParams`
were not properly initialized, and inspecting them raised errors of
missing `_modules`. After this PR, inspecting and using the copies
works.
Test Plan:
```
python test/test_quantization.py TestStaticQuantizedModule.test_linear_api
```
Imported from OSS
Reviewed By: jerryzh168
Differential Revision:
D30702667
fbshipit-source-id:
38c26d1e72663416eeb989985b77ffc2052c12b9
Ivan Kobzarev [Thu, 2 Sep 2021 12:27:59 +0000 (05:27 -0700)]
[jit] shape propagation for prepack (#63585)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/63585
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision:
D30428905
Pulled By: IvanKobzarev
fbshipit-source-id:
c18f6605a69b2e000bdf14a23e637c5a1c2ec64c
Michael Dagitses [Thu, 2 Sep 2021 11:04:59 +0000 (04:04 -0700)]
extract TestAutogradComplex into its own test file (#63400)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63400
This is the first step to break up test_autograd.py for #63205.
Test Plan: Imported from OSS
Reviewed By: albanD
Differential Revision:
D30541499
Pulled By: dagitses
fbshipit-source-id:
8d9d32007938b9eade0e88f95a6a3190e7e2ef01
Michael Dagitses [Thu, 2 Sep 2021 11:04:59 +0000 (04:04 -0700)]
require that `TARGET_DET_LIST` is sorted (and sort it here) (#64102)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64102
We sort this list so that we may add comments to indicate the absence
of a file right where that file would need to be put. This makes it
difficult to wrongly add such a file.
The sorting itself was done programmatically to ensure that no entries
were inadvertently removed.
I printed the sorted list with:
```
for p in sorted(TARGET_DET_LIST):
print(f' "{p}",')
```
Then copied it back into the file.
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision:
D30625076
Pulled By: dagitses
fbshipit-source-id:
cf36fcb3e53e274b76d1f4aae83da1f53c03f9ed
Nicolas Hug [Thu, 2 Sep 2021 10:48:44 +0000 (03:48 -0700)]
Fix list() and help() torchhub functions for Windows (#63773)
Summary:
This PR Fixes the help() and list() torchhub functions which were probably failing for Windows since the `/` OS separator was hardcoded.
Before merging this I need to double check whether the CI actually runs the corresponding tests on Windows or not
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63773
Reviewed By: zou3519
Differential Revision:
D30695664
Pulled By: NicolasHug
fbshipit-source-id:
fac328163fd05db804a8186ae28f22b3cc3a6404
Nicolas Hug [Thu, 2 Sep 2021 10:46:59 +0000 (03:46 -0700)]
Remove outdated comment in hub.py (#63757)
Summary:
This PR removes an outdated comment about Python2 that was orginally introduced in https://github.com/pytorch/pytorch/pull/25083/files. The code has changed since then, but the comment wasn't removed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63757
Reviewed By: zou3519
Differential Revision:
D30695656
Pulled By: NicolasHug
fbshipit-source-id:
431cf414588b9e5a1ad6acdae724ff5af1b16971
Nicolas Hug [Thu, 2 Sep 2021 10:45:06 +0000 (03:45 -0700)]
Update hub.load() signature to avoid polluting kwargs param (#63755)
Summary:
This PR addresses an old comment about Python2 EOL, directly putting some parameters in the function signature instead of in a `**kargs` dict.
I believe the changes are fully backward compatible.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63755
Reviewed By: zou3519
Differential Revision:
D30695634
Pulled By: NicolasHug
fbshipit-source-id:
398f347c5a04bfb58e77e46773a869cb9d0eb225
Kefei Lu [Thu, 2 Sep 2021 08:17:56 +0000 (01:17 -0700)]
Fix TRTModule not adding outputs in order (#64418)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64418
In T99368564, we found that when running TRT lowered module, the output tensors are out-of-order, as compared to the output from the original, non-lowered module. It turns out that in `TRTModule.forward()`, we cannot rely on `ICudaEngine` bindings natural order indices to create the output tensors, but rather, we should explicitly construct the output tensor from the bindings' names, in an ordered that we supply.
Test Plan:
* Arc lint
* Run CI/sandcastle tests
* Run GPU lowering using commands and code changes in
D30171741 and ensure we don't observe out-of-order outputs
Reviewed By: yinghai
Differential Revision:
D30693545
fbshipit-source-id:
32a894ceeb148fcf4e8d279be3835c7d1f1aa2ba
Kushashwa Ravi Shrimali [Thu, 2 Sep 2021 08:08:53 +0000 (01:08 -0700)]
Port `gather` to structured kernel (#63312)
Summary:
Will add a description once this is ready for review.
cc: ysiraichi ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63312
Reviewed By: iramazanli
Differential Revision:
D30597447
Pulled By: ezyang
fbshipit-source-id:
d36e59835c2f4b38e286032dd2a1111a7e16b7e5
Pavel Belevich [Thu, 2 Sep 2021 07:57:39 +0000 (00:57 -0700)]
Replace std::unordered_map<c10::Device, c10::Device> with DeviceMap (#64393)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64393
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse agolynski SciPioneer H-Huang mrzzd cbalioglu gcramer23
Test Plan: Imported from OSS
Reviewed By: rohan-varma
Differential Revision:
D30708384
Pulled By: pbelevich
fbshipit-source-id:
1c565727e4f09cd9e560874dd90aa403470b4a97
Chen Lai [Thu, 2 Sep 2021 07:50:40 +0000 (00:50 -0700)]
[PyTorch Edge] Support default args with out arg, flag off (#63540)
Summary:
1. Allow consuming operators with defaults arguments and out arguments. Flag is off to keep the same behavior as v6, in pr 63651, turn on the flag.
2. Add two unittests to cover this type of operators.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63540
ghstack-source-id:
137211562
Test Plan:
```
caffe2/test/cpp/jit:jit - LiteInterpreterTest.DefaultArgsWithOutArg
caffe2/test/cpp/jit:jit - LiteInterpreterTest.DefaultArgsPinvWithOutArg
```
Reviewed By: raziel, iseeyuan, tugsbayasgalan
Differential Revision:
D30414156
fbshipit-source-id:
0f3a219a22aee10ac53184cbd95940726c459d1f
Edward Yang [Thu, 2 Sep 2021 07:48:03 +0000 (00:48 -0700)]
Remove unnecessary resize_output (#64272)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64272
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Reviewed By: H-Huang, bdhirsh
Differential Revision:
D30686941
Pulled By: ezyang
fbshipit-source-id:
de60e6f1115648f8cf7daaa1e652594fe8b06742
Shirong Wu [Thu, 2 Sep 2021 05:09:42 +0000 (22:09 -0700)]
Move graph util to fx2trt (#64064)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64064
Move original util in torch2trt to fx2trt dir since torch2trt is gonne be deprecated. This is a follow up diff for
D30379124
Test Plan: manual
Reviewed By: yinghai, mikekgfb
Differential Revision:
D30591687
fbshipit-source-id:
ae0e59dfbc2d2e2aa4f3ccea7cff2291c7deb388
Edward Yang [Thu, 2 Sep 2021 04:48:36 +0000 (21:48 -0700)]
Add a warning about DataLoader num_workers > 0 "memory leak" (#64337)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64337
See https://github.com/pytorch/pytorch/issues/13246
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Reviewed By: H-Huang
Differential Revision:
D30690320
Pulled By: ezyang
fbshipit-source-id:
2751aca05a94e63d25162599f458855988516fad
Rohan Varma [Thu, 2 Sep 2021 04:07:01 +0000 (21:07 -0700)]
[Dist CI] Move rest of distributed tests to their own CI job (#64253)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64253
Follow up to
D30496178 (https://github.com/pytorch/pytorch/commit/
f4aff3a346a0525e37d6071f318f7a4c54d5e1fb) to move the rest of distributed tests to their own jobs for Linux GHA.
ghstack-source-id:
137233785
Test Plan: CI
Reviewed By: walterddr
Differential Revision:
D30662999
fbshipit-source-id:
f7cfbc0d1223aca52120f17f9da987d70fda8de6
Rohan Varma [Thu, 2 Sep 2021 01:12:02 +0000 (18:12 -0700)]
[DDP] Log num threads (#64072)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64072
Log gloo threads to DDP logging.
ghstack-source-id:
137119480
Test Plan: CI
Reviewed By: mrshenli
Differential Revision:
D30596083
fbshipit-source-id:
2b4f6e762cb5d850be6056bcc5922029a1af3c91
Zeina Migeed [Thu, 2 Sep 2021 01:04:19 +0000 (18:04 -0700)]
add documentation to shape inference algorithm (#64312)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64312
Test Plan: Imported from OSS
Reviewed By: zou3519
Differential Revision:
D30709254
Pulled By: migeed-z
fbshipit-source-id:
3297d26fe6727c5b9ca176625b1683d787f59659
Yi Wang [Thu, 2 Sep 2021 00:32:39 +0000 (17:32 -0700)]
[DDP Comm Hook] Add debugging communication hooks to ddp_comm_hooks.rst (#64352)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64352
as title
ghstack-source-id:
137246253
Test Plan: N/A
Reviewed By: rohan-varma
Differential Revision:
D30694089
fbshipit-source-id:
a78110b11d59bb0718f43c99ede23f2fd8ab21d0
Yi Wang [Thu, 2 Sep 2021 00:32:39 +0000 (17:32 -0700)]
[DDP Comm Hook] Create a noop hook for performance debugging (#64344)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64344
As title.
Additionally, avoid using numpy array in test_ddp_hooks.py.
ghstack-source-id:
137170449
Test Plan: buck test mode/dev-nosan caffe2/test/distributed/algorithms/ddp_comm_hooks:test_ddp_hooks -- test_ddp_comm_hook_noop_hook
Reviewed By: rohan-varma
Differential Revision:
D30693220
fbshipit-source-id:
e17f0d1c6198863cf20a53566f586a6bff602522
Rohan Varma [Thu, 2 Sep 2021 00:04:37 +0000 (17:04 -0700)]
[DDP] Add more logging iterations (#64071)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64071
Adding more logging iterations to get additional data.
ghstack-source-id:
137119476
Test Plan: CI
Reviewed By: mrshenli
Differential Revision:
D30579367
fbshipit-source-id:
57195266ada5e5926f0d8eaf4fb4e01dc98924d7
Rohan Varma [Wed, 1 Sep 2021 23:25:00 +0000 (16:25 -0700)]
Fix incorrect DDP test (#64074)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64074
Previous PR https://github.com/pytorch/pytorch/pull/63831 did not actually test the error in https://github.com/pytorch/pytorch/issues/63812. Introduce a test
directly from the repro that simulates it.
ghstack-source-id:
137171460
Test Plan: CI
Reviewed By: SciPioneer
Differential Revision:
D30569719
fbshipit-source-id:
fd61250ef6d291c093607663d91d6d2cb5574eb7
Rohan Varma [Wed, 1 Sep 2021 23:21:31 +0000 (16:21 -0700)]
[c10d] Prefer use of torch_check (#63928)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63928
throw std::invalid_argument results in not getting stacktraces with
TORCH_SHOW_CPP_STACKTRACES=1, so instead prefer torch_check here.
ghstack-source-id:
137135328
Test Plan: CI
Reviewed By: mrshenli
Differential Revision:
D30533955
fbshipit-source-id:
33e5bf4f449e3043dec68da93f8022f6624d9675
anjali411 [Wed, 1 Sep 2021 23:11:38 +0000 (16:11 -0700)]
Add fast path for addmm when the inputs are conjugate (#59380)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/59380
Test Plan: Imported from OSS
Reviewed By: ngimel
Differential Revision:
D28898374
Pulled By: anjali411
fbshipit-source-id:
eab0e64d37bb57c18b54cabb8e5c00666338ba04
Yi Wang [Wed, 1 Sep 2021 23:09:46 +0000 (16:09 -0700)]
[DDP Comm Hook] Add bf16 gradient compression to ddp_comm_hooks.rst (#64346)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64346
as title
ghstack-source-id:
137170288
Test Plan: N/A
Reviewed By: rohan-varma
Differential Revision:
D30693513
fbshipit-source-id:
8c64b8404ff3b0322e1bbbd93f6ef051ea91307d
Jerry Zhang [Wed, 1 Sep 2021 22:48:54 +0000 (15:48 -0700)]
[quant][graphmode][fx] Add fbgemm backend_config_dict (#64288)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64288
This is just to setup the file structure and unblock experimentation.
The format for backend_config_dict will change in the future
Test Plan:
python test/test_quantization.py TestQuantizeFx
python test/test_quantization.py TestQuantizeFxOps
Imported from OSS
Reviewed By: zou3519
Differential Revision:
D30699457
fbshipit-source-id:
28211a4def05d34757850c045a36e311f54760fe
Santiago Castro [Wed, 1 Sep 2021 22:18:14 +0000 (15:18 -0700)]
Make datasets in `ConcatDataset` not need to be sized (#64114)
Summary:
`datasets` needs to be iterable, but also sized because the length is checked. But immediately after it's converted to a list. By changing the order of these 2 lines, it doesn't need to be sized anymore.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64114
Reviewed By: H-Huang
Differential Revision:
D30641480
Pulled By: ejguan
fbshipit-source-id:
7e16548c2123afa65b83845f9929271fa07fe1e8
Richard Zou [Wed, 1 Sep 2021 22:12:05 +0000 (15:12 -0700)]
Restore LayerNorm numerics test (#64385)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64385
It was deleted in https://github.com/pytorch/pytorch/pull/63276.
The numerics test was meant to check LayerNorm behavior on large inputs,
but we deleted it without realizing that.
Test Plan: - wait for tests.
Reviewed By: ngimel
Differential Revision:
D30702950
Pulled By: zou3519
fbshipit-source-id:
a480e26c45ec38fb628938b70416cdb22d976a46
Jerry Zhang [Wed, 1 Sep 2021 21:56:14 +0000 (14:56 -0700)]
[quant][graphmode][api] Add backend_config_dict to prepare_fx api (#64135)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64135
We want to start aligning the api with the design in https://github.com/pytorch/pytorch/wiki/Extending-PyTorch-Quantization-to-Custom-Backends
We plan to gradually move things from `prepare_custom_config_dict` and `convert_custom_config_dict`
to `backend_config_dict` and allow custom backend developer to define their own way of quantizing operators.
Test Plan:
python test/test_quantization.py TestQuantizeFx
python test/test_quantization.py TestQuantizeFxOps
Imported from OSS
Reviewed By: zou3519
Differential Revision:
D30699456
fbshipit-source-id:
e3c068da8d3da2270f57719f7159cc71cafa8598
zhouzhuojie [Wed, 1 Sep 2021 21:53:25 +0000 (14:53 -0700)]
Silent rm error for sccache log file (#64388)
Summary:
Sample reporting from dr.ci
![image](https://user-images.githubusercontent.com/658840/
131724645-
75afa04f-7554-4674-8e7c-
cf139c84d994.png)
The `rm` command is not actually running into problems, just need to silent the console output.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64388
Reviewed By: walterddr, malfet, seemethere
Differential Revision:
D30704439
Pulled By: zhouzhuojie
fbshipit-source-id:
ecd35531decf05b75cef30d08d46635f81112f67
Yuchen Huang [Wed, 1 Sep 2021 21:48:00 +0000 (14:48 -0700)]
[xplat][metal] Add getters and setters for ivars in Conv2dOpContext (#57395)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/57395
As title
ghstack-source-id:
137223806
(Note: this ignores all push blocking failures!)
Test Plan:
### Lib Build
- `buck build caffe2:aten_metal_prepack`
### Integration Test
- `arc focus2 pp-ops -a ModelRunner`
- Click "Test Person/Hair Segmentation Model"
{
F612831435}
- Image Classification Demo
{
F614144868}
Reviewed By: xta0
Differential Revision:
D28132020
fbshipit-source-id:
73560263a9d14e9ecfa39c69deb158a2ed8cb179
Meghan Lele [Wed, 1 Sep 2021 21:24:54 +0000 (14:24 -0700)]
[structured] Preserve computed elements from meta func to impl (#61746)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61746
**Summary**
This commit introduces a new feature for structured kernels that allows
kernels to declare quantities as "precomputed" in
`native_functions.yaml`, compute them once in the `meta` function and
reuse them again in the `impl`. The names and types of these quantities
are used to generate code for a struct containing them that the `meta`
function must return. In the case of a handful of surveyed kernels
(`all,`, `any`, `avg_pool2d`), these quantities that are used both in
the `meta` and `impl` have the same meaning as certain kernel arguments
and in fact supersede them. Accordingly, the correspondence between a
kernel argument and the precomputed elements that supersede it is also
captured in `native_functions.yaml`. This information is used to unpack
the struct returned by `meta` and pass its contents correctly to the
`impl` function.
The primary goal is to avoid recompute and enhance developer experience
(e.g. sometimes people can forget to compute these elements while
porting a kernel).
Test Plan: Imported from OSS
Reviewed By: tugsbayasgalan
Differential Revision:
D30407831
Pulled By: SplitInfinity
fbshipit-source-id:
00975525ea373721fe52d06f75cd4ac91f3dc556
Mike Iovine [Wed, 1 Sep 2021 21:19:21 +0000 (14:19 -0700)]
[Static Runtime] Make per-op latency readable by FAI-PEP (#64315)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64315
Add a new flag `generate_ai_pep_output` to `StaticRuntime::benchmark`. If set, produces per-op-kind average total latency in milliseconds in a JSON format recognized by [Facebook AI performance evaluation platform (FAI-PEP)](https://github.com/facebook/FAI-PEP).
This is useful for observing the impact of changes that make a big difference for a specific op, but do not affect the overall SR latency by more than a few percent.
Reviewed By: hlu1
Differential Revision:
D30679352
fbshipit-source-id:
c847fa6ea20774aaf1e7949b11db4421d1f70b7e
Salil Desai [Wed, 1 Sep 2021 21:08:02 +0000 (14:08 -0700)]
Update optimize_for_mobile to preserve node's debug information (#63106)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63106
Propagate debug info to the re-written nodes in the graph.
Test Plan:
- Clone open source repo and build
- ``` python3 test/test_jit.py TestOptimizeForMobilePreserveDebugInfo ```
- Tests pass
Reviewed By: kimishpatel
Differential Revision:
D28654659
fbshipit-source-id:
2d7c87f2fb95a3be53246375f35639bbd97c237e
David Reiss [Wed, 1 Sep 2021 20:41:37 +0000 (13:41 -0700)]
Break up "@generated" string so Phabricator shows changes
Summary: Created from CodeHub with https://fburl.com/edit-in-codehub
Test Plan:
CI
Sandcastle run
Reviewed By: larryliu0820
Differential Revision:
D30701781
fbshipit-source-id:
3acab8b65a327c4ec7da90bc855ecf02f801c40a
Alban Desmaison [Wed, 1 Sep 2021 20:34:48 +0000 (13:34 -0700)]
Add forward AD support for custom Functions (#64061)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64061
Test Plan: Imported from OSS
Reviewed By: soulitzer
Differential Revision:
D30640868
Pulled By: albanD
fbshipit-source-id:
b0e6610430a879074d6d5306443772fc154b431f
Tanvir Zaman [Wed, 1 Sep 2021 20:31:45 +0000 (13:31 -0700)]
Fix bytes_written and bytes_read (#64244)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64244
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64040
In operator cost inference functions, in many places we are using sizeof(x.data_type()). Since data_type() returns a 32 bit integer from [this enum](https://www.internalfb.com/code/fbsource/[
15e7ffe4073cf08c61077c7c24a4839504b964a2]/fbcode/caffe2/caffe2/proto/caffe2.proto?lines=20), we are basically always getting 4 for sizeof(x.data_type()) no matter what actual data type x has. Big thanks to Jack Langman for specifically pointing to this bug.
We would instead use the size in bytes based on actual data type.
Test Plan:
Added unit tests BatchMatMulMemCostTest:
buck test //caffe2/caffe2/fb/fbgemm:batch_matmul_op_test -- BatchMatMulMemCostTest
Extended existing unit test test_columnwise_concat for different data types:
buck test //caffe2/caffe2/python/operator_test:concat_op_cost_test -- test_columnwise_concat
Reviewed By: CrazySherman
Differential Revision:
D30656698
fbshipit-source-id:
d42c0c9a0c5b0ddc5dba39e4994f1f85a5e618bf
Scott Wolchok [Wed, 1 Sep 2021 20:24:11 +0000 (13:24 -0700)]
[Caffe2] Create fewer strings during argument fetching (#64285)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64285
With C++14 heterogeneous ordered container lookup, it is no longer necessary to create a `std::string` in order to look up elements of a `CaffeMap` keyed by std::string. Accordingly, this diff reworks the argument-getting operator functions to avoid that in favor of `c10::string_view`.
ghstack-source-id:
137139818
ghstack-source-id:
137139818
Test Plan: buildsizebot iOS apps -- code size win. less strings is probably marginally good for perf but this only happens at setup time anyway.
Reviewed By: dzhulgakov
Differential Revision:
D26826676
fbshipit-source-id:
ee653b14dc2c528bae8c90f0fc6a7a419cbca1d6
Kimish Patel [Wed, 1 Sep 2021 19:38:39 +0000 (12:38 -0700)]
Back out "Revert
D30327514: [Pytorch lite predictor] Use KinetoEdgeCPUProfiler for operator profiling." (#64307)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64307
Original commit changeset:
0b2aa7c57d08
Restores original changes.
This diff changes the way operator profiling is done in lite predictor
benchmarking binary.
Instead of using custom callbacks it uses KinetoEdgeCPUProfiler to profile
events and then generate operator level metric from it.
Since KinetoEvents do not contain cpu clock time, now we report only wallclock
time.
This unifies various profiling effort that we have for benchmarking purpose. In
production we will still use observer based mechanism, but the advantage of
using kineto profiler is that we get few other things for free, such as:
chrome trace generation.
operator level memory profiling (to be added)
flop counts (to be added)
Furthermore possible we can use python post processing script to parse chrome
trace and generate output similar to torch.profiler. (To be done)
Furthermore removes some tests from test_lite_interpreter.cpp which were testing module hierarchy in debug info. They should be covered by test_mobile_profiler.cpp.
Test Plan:
aibench run
Model without debug info:
https://www.internalfb.com/intern/aibench/details/
219598441154763
Model with debug info and --print_module_info true (see Operator summary has now module hierarchy information).
https://www.internalfb.com/intern/aibench/details/
617154236292985
Reviewed By: raziel
Differential Revision:
D30680354
fbshipit-source-id:
b6ba0d59c510c13d13d9935b1d8051cc82ffa4e9
Rohan Varma [Wed, 1 Sep 2021 19:28:23 +0000 (12:28 -0700)]
Add a record scope around autograd::engine::evaluate_function (#63619)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63619
Adds a RECORD_FUNCTION with the function that is being valuate as part
of backwards execution. This has been useful in picking up some operations
in the backwards pass that otherwise would not show up, for example custom cpp
functions that use custom C++ code.
ghstack-source-id:
137041723
Test Plan:
CI
benchmark:
buck run mode/opt //scripts/rvarm1/ddp:bench
Reviewed By: albanD
Differential Revision:
D30439492
fbshipit-source-id:
955917770cdf2a2edb0303223ace710b668ba388
Patrick Kan [Wed, 1 Sep 2021 19:20:50 +0000 (12:20 -0700)]
[Bootcamp] Include both python unittest and parser parameters in --help and -h flag (#64297)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/45945
Creates a new thread to run -h or --help with unittest.main if the help flag is present, and keeps the add_help default for parameters.
Includes both python unittest and parser parameters in --help and -h flag and will remain up to date since both messages are displayed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64297
Test Plan:
Imported from GitHub
`python test/test_spectral_ops.py --help`
Output:
```
% python test/test_spectral_ops.py --help
usage: test_spectral_ops.py [-h] [-v] [-q] [--locals] [-f] [-c] [-b] [-k TESTNAMEPATTERNS] [tests [tests ...]]
positional arguments:
tests a list of any number of test modules, classes and test methods.
optional arguments:
-h, --help show this help message and exit
-v, --verbose Verbose output
-q, --quiet Quiet output
--locals Show local variables in tracebacks
-f, --failfast Stop on first fail or error
-c, --catch Catch Ctrl-C and display results so far
-b, --buffer Buffer stdout and stderr during tests
-k TESTNAMEPATTERNS Only run tests which match the given substring
Examples:
test_spectral_ops.py - run default set of tests
test_spectral_ops.py MyTestSuite - run suite 'MyTestSuite'
test_spectral_ops.py MyTestCase.testSomething - run MyTestCase.testSomething
test_spectral_ops.py MyTestCase - run all 'test*' test methods
in MyTestCase
usage: test_spectral_ops.py [-h] [--subprocess] [--seed SEED] [--accept] [--jit_executor JIT_EXECUTOR] [--repeat REPEAT]
[--test_bailouts] [--save-xml [SAVE_XML]] [--discover-tests] [--log-suffix LOG_SUFFIX]
[--run-parallel RUN_PARALLEL] [--import-slow-tests [IMPORT_SLOW_TESTS]]
[--import-disabled-tests [IMPORT_DISABLED_TESTS]]
optional arguments:
-h, --help show this help message and exit
--subprocess whether to run each test in a subprocess
--seed SEED
--accept
--jit_executor JIT_EXECUTOR
--repeat REPEAT
--test_bailouts
--save-xml [SAVE_XML]
--discover-tests
--log-suffix LOG_SUFFIX
--run-parallel RUN_PARALLEL
--import-slow-tests [IMPORT_SLOW_TESTS]
--import-disabled-tests [IMPORT_DISABLED_TESTS]
```
Also ran some other tests to make sure tests still worked, and other tests with --help or -h flag
Reviewed By: seemethere
Differential Revision:
D30677776
Pulled By: PatrickKan
fbshipit-source-id:
eb3d6e3fa677137ec703ec3a23808efb99acc896
Patrick Hu [Wed, 1 Sep 2021 17:49:39 +0000 (10:49 -0700)]
[FX] Fix python code generation for wrapped getattr() with default value (#64271)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64271
Closes #60417
Modified emit_node() in fx/graph.py to generate getattr() call with default value when len(node.args) != 2 instead of accessing the attribute.
Added test_torch_fx_getattr() in test/test_fx.py.
Test Plan:
pytest test/test_fx.py
Imported from OSS
Reviewed By: jamesr66a
Differential Revision:
D30671265
fbshipit-source-id:
f2db9ea47e0cb247547e200684f715aab006c374
Raghavan Raman [Wed, 1 Sep 2021 17:28:02 +0000 (10:28 -0700)]
[nnc] Updated generic error message with info about turning off the fuser (#64316)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64316
Test Plan: Imported from OSS
Reviewed By: bertmaher
Differential Revision:
D30683942
Pulled By: navahgar
fbshipit-source-id:
d86607563672213f99a1436dcf4f5dc28053b713
Xiang Gao [Wed, 1 Sep 2021 17:17:52 +0000 (10:17 -0700)]
Fixes reduction launch config (#64304)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/48573
See also https://github.com/pytorch/pytorch/pull/64194
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64304
Reviewed By: janeyx99
Differential Revision:
D30689600
Pulled By: ngimel
fbshipit-source-id:
bf2103ca177fd3b6e27bc0324b81925234483a29
Kushashwa Ravi Shrimali [Wed, 1 Sep 2021 15:48:25 +0000 (08:48 -0700)]
OpInfo for `nn.functional.layer_norm` (#63276)
Summary:
Please see https://github.com/facebookresearch/functorch/issues/78 and https://github.com/pytorch/pytorch/issues/54261.
Note:
* This PR also adds a reference test inspired by existing tests in `test_nn.py`.
cc: mruberry zou3519
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63276
Reviewed By: ejguan
Differential Revision:
D30452483
Pulled By: zou3519
fbshipit-source-id:
2578d01ca34e031668a41bd284db60c31ae1fba8
Nima Elyasi [Wed, 1 Sep 2021 15:47:44 +0000 (08:47 -0700)]
fix GradBucket.is_last() logic (#63768)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63768
passed number of buckets to GradBucket constructor, to check if index is equal to num_buckets - 1 in the .is_last() function.
Test Plan:
buck test mode/dev-nosan //caffe2/test/distributed/algorithms/ddp_comm_hooks:test_ddp_hooks
test output: https://www.internalfb.com/intern/testinfra/testconsole/testrun/
8162774375985873/
Reviewed By: SciPioneer, mrshenli
Differential Revision:
D30455913
fbshipit-source-id:
8c67ca69cbf191d6e189e09248407eb167bb24b6
Richard Zou [Wed, 1 Sep 2021 14:16:55 +0000 (07:16 -0700)]
Revert
D29699456: [pytorch][PR] Enable Half, BFloat16, and Complex dtypes for coo-coo sparse matmul [CUDA]
Test Plan: revert-hammer
Differential Revision:
D29699456 (https://github.com/pytorch/pytorch/commit/
ad4848565e1d9f4d408c60614f213acb52035181)
Original commit changeset:
407ae53392ac
fbshipit-source-id:
b6c70ba8bb28c0c38de47857030b69792a8470de
James Reed [Wed, 1 Sep 2021 05:20:41 +0000 (22:20 -0700)]
[FX] Rename reduce functions back to their old, public names (#64324)
Summary:
Unfortunately pickle serializes the names of these functions. Also put them under backward-compatibility enforcement.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64324
Test Plan: Local repro https://fb.workplace.com/groups/
3440841732711443/permalink/
4018921611570116/
Reviewed By: SplitInfinity, TailofJune
Differential Revision:
D30684185
Pulled By: jamesr66a
fbshipit-source-id:
900701220155d15115cd0c07cf7774a2891bd04f
Yuchen Huang [Wed, 1 Sep 2021 05:00:11 +0000 (22:00 -0700)]
[Metal][GPU] Enable metal for simulators and fix test failures if possible (#64322)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64322
As title
ghstack-source-id:
137143877
Test Plan:
- `aibench-cli mobile`
- Select iOS -> `y` -> `1` -> `n` -> "--metal_op_test"
- Select all iPhone 6 + iPhone 7 + iPhone 8 and a iPhone X or 11 or 12
```
Benchmark Submitted. Find more details at: https://our.intern.facebook.com/intern/aibench/details/
318120612514604
Benchmark Status:
D10 (https://github.com/pytorch/pytorch/commit/
b8256280ce45f02a7e105d3b3db4a547990e683d)AP-12.0.1: DONE
N71mAP-14.3: DONE
DUMMY latency:
D10 (https://github.com/pytorch/pytorch/commit/
b8256280ce45f02a7e105d3b3db4a547990e683d)AP-12.0.1: 4319.3
N71mAP-14.3: 8868.51
I0831 16:06:27.210558 605277 ClientSingletonManager.cpp:99] Shutting down Manifold ClientSingletonManager
```
Reviewed By: xta0
Differential Revision:
D30147163
fbshipit-source-id:
2de6bbd9bd525e32ca92b2845eb435800855edcc
Michael Carilli [Wed, 1 Sep 2021 04:43:25 +0000 (21:43 -0700)]
[CUDA graphs] hotfix for test_graph_ (#64339)
Summary:
Graphed workloads that try to capture a full backward pass must do warmup on a non-default stream. If warmup happens on the default stream, AccumulateGrad functions might tag themselves to run on the default stream, and therefore won't be capturable.
ngimel and I suspect some test_cuda.py tests run with the default stream as the ambient stream, which breaks `test_graph_grad_scaling` because `test_graph_grad_scaling` does warmup on the ambient stream _assuming_ the ambient stream is a non-default stream.
This PR explicitly sets a side stream for the warmup in `test_graph_grad_scaling`, which is what I should have done all along because it's what the new documentation recommends.
I pushed the PR branch straight to the main pytorch repo because we need to run ci-all on it, and I'm not sure what the requirements are these days.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64339
Reviewed By: mruberry
Differential Revision:
D30690711
Pulled By: ngimel
fbshipit-source-id:
91ad75f46a11f311e25bc468ea184e22acdcc25a
gmagogsfm [Wed, 1 Sep 2021 04:27:46 +0000 (21:27 -0700)]
Remove outdated warning about RecursiveScriptModule not being copiable (#64085)
Summary:
RecursiveScriptModule has its customized `__copy__` and `__deepcopy__` defined. The warning/error that says it is not copiable is outdated
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64085
Reviewed By: rohan-varma
Differential Revision:
D30598623
Pulled By: gmagogsfm
fbshipit-source-id:
0701d8617f42d818bc7b88244caee4cd47fbe976
Mikhail Zolotukhin [Wed, 1 Sep 2021 03:27:44 +0000 (20:27 -0700)]
[TensorExpr] Wrap error messages with buildErrorMessage call. (#64330)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64330
Test Plan: Imported from OSS
Reviewed By: bertmaher
Differential Revision:
D30687226
Pulled By: ZolotukhinM
fbshipit-source-id:
ade1be2ad6847c6afbba60307ef854696821b4e3
Pritam Damania [Wed, 1 Sep 2021 03:19:55 +0000 (20:19 -0700)]
Fix bug in ShardedTensorMetadata serde. (#63902)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63902
The 'memory_format' field was not being serialized correctly and used
the same encoding for different fields.
ghstack-source-id:
137142406
Test Plan: waitforbuildbot
Reviewed By: bowangbj
Differential Revision:
D30527324
fbshipit-source-id:
f0f223e2d660ef6e4abae9649d9992acc36e1278
Pavel Belevich [Wed, 1 Sep 2021 03:14:08 +0000 (20:14 -0700)]
Delete some dead code from RRefMessageBase (#64298)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64298
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse agolynski SciPioneer H-Huang mrzzd cbalioglu gcramer23
Test Plan: Imported from OSS
Reviewed By: rohan-varma
Differential Revision:
D30676702
Pulled By: pbelevich
fbshipit-source-id:
77dbc0f8064c3518376454ff573d45ed0274956b
Matti Picus [Wed, 1 Sep 2021 01:54:44 +0000 (18:54 -0700)]
disallow empty named dims list to flatten(names, name) (#61953)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/61137 by raising an error if an empty tuple is passed in for the names:
```
>>> torch.empty((2, 3), names=['a', 'b']).flatten((), 'abc')
RuntimeError: flatten(tensor, dims, out_dim): dims cannot be empty
```
or from the original issue:
```
>>> torch.empty((2, 3)).flatten((), 'abc')
RuntimeError: flatten(tensor, dims, out_dim): dims cannot be empty
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61953
Reviewed By: iramazanli
Differential Revision:
D30574571
Pulled By: malfet
fbshipit-source-id:
e606e84458a8dd66e5da6d0eb1a260f37b4ce91b
Scott Wolchok [Wed, 1 Sep 2021 01:22:23 +0000 (18:22 -0700)]
[caffe2][easy] Save heap allocation in ConcatOp (#63529)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/63529
Output() takes an IntArrayRef, so we can just use a std::initializer_list (stack-allocated array) instead of std::vector here.
ghstack-source-id:
137085908
Test Plan: existing CI
Reviewed By: mruberry
Differential Revision:
D29687400
fbshipit-source-id:
9f2a7c6679f2552c098bb1bf7befaca18e0e5d4d
Edward Yang [Wed, 1 Sep 2021 00:55:23 +0000 (17:55 -0700)]
Convert mul to use opmath_gpu_kernel_with_scalars (#64019)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64019
Note that previously the functor operated on scalar_t and
this modifies it to operate on opmath_t, but this is not
a problem as half precision was implemented by performing the
compute in float anyway.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Reviewed By: ngimel
Differential Revision:
D30575282
Pulled By: ezyang
fbshipit-source-id:
cc6900ef996e755740afe48f9cb4d0366858dd47
soulitzer [Wed, 1 Sep 2021 00:51:55 +0000 (17:51 -0700)]
Use the correct overloaded name to skip boxed autograd not implemented kernel registration (#64182)
Summary:
Some internal use_count tests are failing for `dequantize_self` because we only compare the skip list with the base name `dequantize` when we should be comparing with the full name including the overload
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64182
Reviewed By: albanD
Differential Revision:
D30639909
Pulled By: soulitzer
fbshipit-source-id:
d4d22dd1a5c8f7180251ce7739830764cce6f151
Ray Peng [Wed, 1 Sep 2021 00:45:50 +0000 (17:45 -0700)]
[Static Runtime] Out version for softmax (#64243)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/64243
Test Plan:
```
> buck run //caffe2/benchmarks/static_runtime:static_runtime_cpptest -- --v=1
...
V0830 16:35:22.524479 613839 impl.cpp:1410] Switch to out variant for node: %5 : Tensor = aten::softmax(%a.1, %dim.1, %dtype.1)
...
[ OK ] StaticRuntime.IndividualOps_Softmax (803 ms)
```
Reviewed By: hlu1
Differential Revision:
D30656149
fbshipit-source-id:
115b7b4a75448fd6a5c526808080ca9a4251302c
Eli Uriegas [Wed, 1 Sep 2021 00:38:42 +0000 (17:38 -0700)]
.circleci: Remove already migrated CUDA configs (#64231)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64231
This migrates over the CUDA 11.1 and CUDA 10.2 configs that we had
previously migrated to GHA
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
cc ezyang seemethere malfet walterddr lg20987 pytorch/pytorch-dev-infra
Test Plan: Imported from OSS
Reviewed By: zhouzhuojie
Differential Revision:
D30683811
Pulled By: seemethere
fbshipit-source-id:
71b0761461557d871c26eb02f665a2e4d9b1d9fb
Eli Uriegas [Wed, 1 Sep 2021 00:38:42 +0000 (17:38 -0700)]
.github: Consolidate linux setup / teardown (#64229)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64229
Consolidates linux setup / teardown into easy to use jinja2 macros
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
cc ezyang seemethere malfet walterddr lg20987 pytorch/pytorch-dev-infra
Test Plan: Imported from OSS
Reviewed By: zhouzhuojie, driazati
Differential Revision:
D30683810
Pulled By: seemethere
fbshipit-source-id:
2578630df3e212fb79392a699090553baef44cc2
Nikita Shulga [Wed, 1 Sep 2021 00:33:11 +0000 (17:33 -0700)]
Add ciflow-tracking issue to pytorch-probot (#64125)
Summary:
Doesn't do anything yet...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64125
Reviewed By: zhouzhuojie
Differential Revision:
D30620283
Pulled By: malfet
fbshipit-source-id:
91869d35c1b70a55e32261d2c32fb0136ec33960