review.tizen.org Git - platform/upstream/pytorch.git/log

Skipping two c10d tests only if there are multi-GPUs (#14860)

Summary:
Otherwise, these tests will fail, even though there are never meant to run on single GPU machines.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14860

Differential Revision: D13369060

Pulled By: teng-li

fbshipit-source-id: 8a637a6d57335491ba8602cd09927700b2bbf8a0

Move TensorOptions, DefaultTensorOptions to c10

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14746

Reviewed By: ezyang

Differential Revision: D13318644

fbshipit-source-id: b703d7dc67e75d9e9571c80d62a100c5fc4e84df

Switch Int8MaxPool operator to QNNPACK (#14832)

Summary:
1.6-2.4X speedup on ARM when compiled with gcc
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14832

Differential Revision: D13358160

Pulled By: Maratyszcza

fbshipit-source-id: 39e9791886fac62650bb53a9df341889f0bb5d49

collect_env.py: get conda magma and mkl information (#14854)

Summary:
Fixes #12371
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14854

Differential Revision: D13363635

Pulled By: zou3519

fbshipit-source-id: f8b5d05038bf5ce451399dfeed558ae298178128

Add LogSigmoid support in ONNX symbolic (#14830)

Summary:
Add LogSigmoid:

torch.LogSigmoid(x) = onnx.Log(onnx.Sigmoid(x))
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14830

Differential Revision: D13353891

Pulled By: zrphercule

fbshipit-source-id: bf456170b9e6c4edad07b3333cd5797f8e0fa97f

Kill GPU memory logs in normal runs (#14838)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14838

The GPU memory tracking logs are incredibly annoying and merely serve
to pollute output. I `VLOG(1)`ed them. Hopefully, this is non-controversial.

Reviewed By: kuttas

Differential Revision: D13343290

fbshipit-source-id: b3cae99346c97b66e97ea660061e15dc5c99b9fc

Stop inserting static casts in Hipify (#14853)

Summary:
Latest hcc can now properly cast to correct type internally, so there is no need to insert static_cast in hipify scripts anymore.
However the hcc included in the latest ROCm release (1.9.2) doesn't have this fix, so leaving a flag to continue doing static_cast for those using the official ROCm releases.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14853

Differential Revision: D13363171

Pulled By: bddppq

fbshipit-source-id: a36476a8511222ff3c933d31788e8a0ffb04f5ca

Tensor construction codemod - 3/3 (#14835)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14835

Codemod generated with clangr shard mode, 25 files per diff,
motivation: https://github.com/pytorch/pytorch/pull/12407

Reviewed By: bddppq

Differential Revision: D13335184

fbshipit-source-id: 26d8247e16b30bdff045530034af9b72c76d066f

Tensor construction codemod - 1/3 (#14828)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14828

Codemod generated with clangr shard mode, 25 files per diff,
motivation: https://github.com/pytorch/pytorch/pull/12407

Reviewed By: bddppq

Differential Revision: D13335160

fbshipit-source-id: a3ae4c5a86bfbdaf2d5aa14e0eef57255e829fd4

Move numa.{h, cc} to c10/util (#14393)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14393

att

Reviewed By: ezyang

Differential Revision: D13205604

fbshipit-source-id: 54166492d31827b0343ed070cc36a825dd86e2ed

Upgrade CI to ROCm 1.9.2 (#14216)

Summary:
Drop custom hcc/hip as the 1.9.2 release should contain the relevant patches therein.

Most notable feature in 1.9.2 is mixed precision support in rocBLAS and MIOpen. These features will be enabled by subsequent PRs.

bddppq ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14216

Differential Revision: D13354294

Pulled By: bddppq

fbshipit-source-id: 2541d4a196af21c9432c1aff7f6e65b572628028

Allow linspace and logspace with steps=1 and start != end like numpy (#14748)

Summary:
`torch.linspace(0, 1, 1)` fails with `RuntimeError: invalid argument 3: invalid number of points at ../aten/src/TH/generic/THTensorMoreMath.cpp:2119`, while `np.linspace(0, 1, 1)` works fine.
Looking at the code, there is even a comment by gchanan asking: "NumPy allows you to pass different points even if n <= 1 -- should we?"
I would say "yes". Currently, I would need to handle the case of `steps == 1` or `steps == 0` separately, making sure to change the `end` when calling `torch.linspace`. This is impractical. If we support `start != end`, there are two possibilities for the result: Either we ensure the first value in the resulting sequence always equals `start`, or we ensure the last value in the resulting sequence always equals `end`. Numpy chose the former, which also allows it to support a boolean `endpoint` flag. I'd say we should follow numpy.

This PR adapts `linspace` and `logspace` to mimic the behavior of numpy, adapts the tests accordingly, and extends the docstrings to make clear what happens when passing `steps=1`.

If you decide against this PR, the error message should become explicit about what I did wrong, and the documentation should be extended to mention this restriction.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14748

Differential Revision: D13356136

Pulled By: ezyang

fbshipit-source-id: db85b8f0a98a5e24b3acd766132ab71c91794a82

(#14580)

Summary:
Removes cast of half to float in torch.sum, with float16 input tensor and
float32 output tensor, instead we cast data when loading input in kernel.

This supposingly would save a kernel launch as well as a full global memory load
on promoted data type (float).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14580

Differential Revision: D13356203

Pulled By: ezyang

fbshipit-source-id: 85e91225b880a65fe3ceb493371b9b36407fdf48

Consistent formatting in losses' docs

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14739

Differential Revision: D13356143

Pulled By: ezyang

fbshipit-source-id: 9ae8316dd8ba6e910247b64cec22db63df10e11c

Add (partial) autodiff support for nll_loss (#14305)

Summary:
Not ready yet, need some comments / help with this. It's good enough for https://github.com/pytorch/xla immediate goals (forward + backward trace fusion), but there are at least two issues with it:

1. If we don't allow it, `test/test_jit.py` fails to cover the change.
2. If we allow the weight to be set, running `test/test_jit.py TestJitGenerated.test_nn_nll_loss` fails with:

```
======================================================================
ERROR: test_nn_nll_loss (__main__.TestJitGenerated)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test/test_jit.py", line 10001, in do_test
    fn, f_args_variable, kwargs_variable, no_grad=no_grad)
  File "test/test_jit.py", line 9360, in check_against_reference
    outputs_test = self.runAndSaveRNG(func, recording_inputs, kwargs)
  File "test/test_jit.py", line 425, in runAndSaveRNG
    results = func(*inputs, **kwargs)
  File "test/test_jit.py", line 9298, in script_fn
    self.assertExportImport(CU.the_method.graph, tensors)
  File "test/test_jit.py", line 415, in assertExportImport
    self.assertExportImportModule(m, inputs)
  File "test/test_jit.py", line 419, in assertExportImportModule
    self.assertEqual(self.runAndSaveRNG(m.forward, inputs),
  File "test/test_jit.py", line 425, in runAndSaveRNG
    results = func(*inputs, **kwargs)
RuntimeError:
arguments for call are not valid:

  for operator aten::nll_loss_backward(Tensor grad_output, Tensor self, Tensor target, Tensor? weight, int reduction, int ignore_index, Tensor total_weight, *, Tensor out) -> Tensor:
  expected a value of type Tensor for argument 'total_weight' but found bool
  <internally-created-node>
  ~ <--- HERE

  for operator aten::nll_loss_backward(Tensor grad_output, Tensor self, Tensor target, Tensor? weight, int reduction, int ignore_index, Tensor total_weight) -> Tensor:
  expected a value of type Tensor for argument 'total_weight' but found bool
  <internally-created-node>
  ~ <--- HERE
for call at:
<internally-created-node>
~ <--- HERE
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14305

Differential Revision: D13356265

Pulled By: ezyang

fbshipit-source-id: 504d783b2d87f923e698a6a4efc0fd9935a94a41

Updating submodules

Reviewed By: yns88

fbshipit-source-id: 2adbb6f97d4b8f067a2538fec855063510b0ca3f

Updating submodules

Reviewed By: yns88

fbshipit-source-id: e0509413215f3b7578b825c52365fec4da625bd5

Fixed MIOpen RNN Segfault issue and enabled RNN test (#14810)

Summary:
This pull request contains changes for:
1. Added MIOpen RNN API miopenGetRNNLayerBiasSize and miopenGetRNNLayerParamSize.
2. Fixed usage of API miopenGetRNNLayerParam.
3. Modifying the RNN test to run using MIOpen engine.

Differential Revision: D13355699

Pulled By: bddppq

fbshipit-source-id: 6f750657f8049c5446eca893880b397804120b69

Export complete subgraph io info when calling onnxGetBackendCompatibility (#14827)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14827

We need to send complete IO info when doing `onnxGetBackendCompatibility` to backend like Glow. Previously we are missing some info because sometimes we generate more than one nodes from one C2 op. This fixes the issue.

Reviewed By: jackm321

Differential Revision: D13352049

fbshipit-source-id: 8d8ac70656a0ac42f3a0ccecad61456a4f3b2435

Fix clip gradient with empty input (#14709)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14709

As titled

Reviewed By: Wakeupbuddy

Differential Revision: D13305554

fbshipit-source-id: 380062d4b0e4f9dc0207a27766cac7b8d05384d5

Remove protobuf dependency in pytorch cmake file. (#14182)

Summary:
Currently, pytorch doesn't dependent on protobuf. So, we don't need to include the protobuf dir in pytorch cmake file.
And if we build caffe2 without custom-protobuf[1], we will have the protobuf mismatched problem.

[1]
https://github.com/pytorch/pytorch/blob/92dbd0219f6fbdb1db105386386ccf92c0758e86/CMakeLists.txt#L65
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14182

Differential Revision: D13356273

Pulled By: ezyang

fbshipit-source-id: 8120c3452d158dc51d70156433d7b9076c6aed47

Optimize images (#14084)

Summary:
This is a PR that [ImgBot](https://imgbot.net/) opened on my fork https://github.com/zasdfgbnm/pytorch/pull/1, I forward it here. ImgBot does lossless compression on images to reduce file size.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14084

Differential Revision: D13356293

Pulled By: ezyang

fbshipit-source-id: 731236d95ad870db8ccb99b03ed306704365242c

Prevent `profile_observer_test` from being run by CPU test (#14168)

Summary:
Fix CMakeLists.txt, so the test for CPU won't run profile_observer_test.cc, as currently it only supports GPU
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14168

Differential Revision: D13356274

Pulled By: ezyang

fbshipit-source-id: 7d105f2e18675e5fab129864958148b0f18d582c

CAFFE2_INCLUDE_DIRS points to invalid path (#14306)

Summary:
I know that including CAFFE2_INCLUDE_DIRS in include headers are not necessary for newer cmakes. But I had this in one of my old projects and **cmake gave me error that "/usr/lib/include" is invalid path**.

It seems like "${_INSTALL_PREFIX}/lib/include" should be changed to "${_INSTALL_PREFIX}/include" as all caffe2 headers are in /include rather than /lib/include/

Please correct me if I am wrong?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14306

Differential Revision: D13356246

Pulled By: ezyang

fbshipit-source-id: e2d5d3c42352e59b245714ad90fd7a9ef48170d7

use "Extension" instead of the unimported "setuptools.Extension" (#14475)

Summary:
use "Extension" instead of the unimported "setuptools.Extension"
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14475

Differential Revision: D13356219

Pulled By: ezyang

fbshipit-source-id: 5a3e7eb73a32d6bf09676efd9eddded5586435cd

generate ATen core files with LF. (#14667)

Summary:
on Windows environment, some ATen core files (Type.h, Tensor.h, TensorMethods.h) are created and it's new line code is CRLF. (maybe enviconment dependant)
therefore, comparing files is failed in generate_outputs()agener917.py and compilation stopped.
this patch generates these files with LF forcibly.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14667

Differential Revision: D13356170

Pulled By: ezyang

fbshipit-source-id: ef8cc3a6cc8bf3c45b78e9eb3df98cf47c0d33bb

Remove outdated css file and refs in cpp conf.py (#14779)

Summary:
pytorch_theme.css is no longer necessary for the cpp or html docs site build. The new theme styles are located at https://github.com/pytorch/pytorch_sphinx_theme. The Lato font is also no longer used in the new theme.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14779

Differential Revision: D13356125

Pulled By: ezyang

fbshipit-source-id: c7635eb7512c7dcaddb9cad596ab3dbc96480144

Fixes for some Windows compiler warnings (#14490)

Summary:
Implement some simple fixes to clean up windows build by fixing compiler warnings. Three main types of warnings were fixes:

1. GCC specific pragmas were changed to not be used on windows.
2. cmake flags that don't exist on windows were removed from windows build
3. Fix a macro that was defined multiple times on Windows.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14490

Differential Revision: D13241988

Pulled By: ezyang

fbshipit-source-id: 38da8354f0e3a3b9c97e33309cdda9fd23c08247

Shut up "address will always evaluate to 'true'" warnings (#14774)

Summary:
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14774

Differential Revision: D13327969

Pulled By: ezyang

fbshipit-source-id: 43380c89eedaaa89467952401b8fd3f5a9ad754a

HIPify less files in PyTorch (#14804)

Summary:
Stacked on #14803
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14804

Differential Revision: D13347986

Pulled By: ezyang

fbshipit-source-id: c93177b4ad51855660d0de36d042bfc542bd4be0

Unify device argument parsing between torch and c10

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14786

Differential Revision: D13334501

Pulled By: bddppq

fbshipit-source-id: ae3536be1fe0dcd6a1552ec93629ecc9554c0d7c

Improve assertion failure message (#14813)

Summary:
See #14554.

I can't figure out how the reported issue can happen. The best next
thing is have more information when this happens again.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14813

Differential Revision: D13351908

Pulled By: pietern

fbshipit-source-id: 61b30fcae2e34da54329d0893ca4921b6ad60f0d

Add FunctionSchema based Operator Registry (#13789)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13789

This enables creation of operators with FunctionSchema and IValue

Reviewed By: smessmer

Differential Revision: D13008791

fbshipit-source-id: 151efc88ac315f4a0ab0171a99774caaf767ef1e

Increase test timeout (#14814)

Summary:
It is possible that some sort of contention causes process scheduling
delays which in turn cause the timeout to *not* be hit.

Increased sleep here will decrease the probability of this happening.

Fixes #14555.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14814

Differential Revision: D13351924

Pulled By: pietern

fbshipit-source-id: 1222cf0855408dfcb79f30f94694c790ee998cf9

Retry test on address already in use error (#14815)

Summary:
Thanks nairbv for the suggestion.

Also see #14589.

Fixes #14703.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14815

Differential Revision: D13351913

Pulled By: pietern

fbshipit-source-id: d11a4152505d0ce15592b13e417bb80551476a61

improve ONNX tests on torch.Linear

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14821

Reviewed By: zrphercule

Differential Revision: D13348773

Pulled By: houseroad

fbshipit-source-id: 611ca6e28f715e5518649c8c16f702ac3433308c

Define THPStorage struct only once (rather than N times) (#14802)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14802

The definetion of THPStorage does not depend on any Real, its macro
defintion is unnecessary, refactor the code so that THPStorage is not macro
defined.

Reviewed By: ezyang

Differential Revision: D13340445

fbshipit-source-id: 343393d0a36c868b9a06eea2ad9b80f5e395e947

File name change for FbgemmI8Depthwise.h and FbgemmI8Depthwise.cc (#14725)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14725

Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/33

Renaming FbgemmI8Depthwise.h to FbgemmI8DepthwiseAvx2.h and FbgemmI8Depthwise.cc to FbgemmI8DepthwiseAvx2.cc since FbgemmI8DepthwiseAvx2.cc will be compiled with avx2 flags

Reviewed By: jianyuh

Differential Revision: D13313898

fbshipit-source-id: a8111eacf3d79a466ce0565bfe5f2f0b200a5c33

Add torch.nn.RReLU support in symbolic (#14781)

Summary:
Now we support exporting torch.nn.RReLU in onnx.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14781

Reviewed By: houseroad

Differential Revision: D13343872

Pulled By: zrphercule

fbshipit-source-id: 1e96b957de4fc2f5ba3959d42329807975419ae3

Move avx2 specific code in different source files (#28)

Summary:
Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/28

Pull Request resolved: https://github.com/pytorch/pytorch/pull/14516

This is the first diff in a series of diffs that will separate out avx2 specific code in separate files. The goal is to compile as little as possible code with avx2 and avx512 compiler flags.

Reviewed By: jianyuh

Differential Revision: D13248376

fbshipit-source-id: 401c2e9d3cd96c420fd08c3efa011febce96ffbb

Validate matching input shapes in Int8Add operator (#14520)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14520

Default engine doesn't support broadcast semantics in Int8Add operator. This patch adds a check that shapes are equivalent.

Reviewed By: bertmaher

Differential Revision: D13250922

fbshipit-source-id: 8526d07723bd9a34d54dee04d121c57f8b33c481

fix stft arg types

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14800

Reviewed By: zou3519

Differential Revision: D13340574

Pulled By: SsnL

fbshipit-source-id: 8b0dbbe299d1a362da0ecc0b1c0dadb2543ded5d

Improve HIPify performance (#14803)

Summary:
```
    Improve performance of pyHIPIFY

    Changes:
    - Pre-compile regexes, don't use regexes when it's not necessary
      (this saves us ~15%)
    - Compile all substitutions for mappings into a single, non-backtracking
      regex using a Trie.  This gives big savings.

    Before, running pyHIPIFY on all files took 15.8s.  Now it takes 3.9s.
```

Stacked on #14769
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14803

Differential Revision: D13342620

Pulled By: ezyang

fbshipit-source-id: 1cfa36b3236bbe24d07080a31cc788a52d740f40

Fix cuda multiprocessing cached memory (#14736)

Summary:
This PR fixes #11422

In the old world of CUDA IPC, when we want to share a tensor T from A to B, we have to share the whole CUDA mem allocation where T's storage sit in. And we casted it to the same type of storage of T's.

This causes problem when two different types of storage got allocated to the same CUDA mem block. When we try to reconstruct the second tensor, it will complain about wrong storage type.

In this PR we reconstruct the storage only (not the entire mem block). However, CUDA only allows one open memHandle once per process, we have to save the device pointer in a global cache so that we can reconstruct tensors as they come.

Thanks a ton to ezyang who helped design the solution and debugged the issue!
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14736

Differential Revision: D13335899

Pulled By: ailzhang

fbshipit-source-id: cad69db392ed6f8fdc2b93a9dc2899f6d378c371

Set and get default dtype (#13748)

Summary:
Replaces the `DefaultTensorOptions` with just a global default dtype that you can set and get like in Python.

Also, calls `set_default_dtype` in the implementation of `torch.set_default_dtype`. Right now these two default values are separate but will always be the same. Should we just bind `set_default_dtype` into Python? I think that might be good to do in a separate PR though.

ezyang gchanan

Also CC colesbury who wanted to do this for ATen for a while? What do you think about it?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13748

Differential Revision: D13340207

Pulled By: goldsborough

fbshipit-source-id: 2689b09eb137fabb3a92d1ad1635782bee9398e8

Switch Int8AveragePool operator to QNNPACK (#14783)

Summary:
2.2-2.9X better performance on ARM when compiled with gcc (same bad perf when compiled with Clang)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14783

Differential Revision: D13332680

Pulled By: Maratyszcza

fbshipit-source-id: 4c1138500c6b3026335e9bfe5f6be43b1ae2cefb

Update magma to 2.4.0 for Windows

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14738

Differential Revision: D13341611

Pulled By: soumith

fbshipit-source-id: 39a49fc60e710cc32a463858c9cee57c182330e2

Unify build_caffe2_amd.py and build_pytorch_amd.py (#14769)

Summary:
I need to preserve ability to HIPify out-of-place files
only, so build_amd.py grows a --out-of-place-only flag.

Stacked on #14757
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14769

Differential Revision: D13340154

Pulled By: ezyang

fbshipit-source-id: 1b855bc79e824ea94517a893236fd2c8ba4cb79d

Default pool() option (#14636)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14636

Add a default CPU option for the pool()

Reviewed By: andrewwdye

Differential Revision: D13281367

fbshipit-source-id: 92dbfce89c900a41731b6d1ff62bb97886c40f77

Storage.clone maintains original device (#14751)

Summary:
Fixes https://github.com/pytorch/pytorch/issues/14673

As pointed out by vishwakftw , the root case of the `deepcopy` issue was that `storage.clone()` would create a new storage in the default device.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14751

Reviewed By: soumith

Differential Revision: D13323061

Pulled By: fmassa

fbshipit-source-id: bfe46ebd78f0b6cd9518c11d09de7849282ed2a2

Updating submodules

Reviewed By: yns88

fbshipit-source-id: 080e0034bd6353420383ac7b476af5a35eaba7c3

Updating submodules

Reviewed By: yns88

fbshipit-source-id: e397238c7c477c4268e2dc89e530776fc89f18f8

include avx512vl to avx512 code path (#14733)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14733

We often also want to use AVX512VL instruction sets.
We already included AVX512F, AVX512DQ.
Skylake also has AVX512BW, AVX512CD we may want to later.

Reviewed By: duc0

Differential Revision: D13317282

fbshipit-source-id: 82c8e401d82d5c3a5452fb4ccb6e5cb88d242bda

Use AT_WARN for warnings in the JIT (#14770)

Summary:
Previously their implementation dispatched to prim::Print, which kept
printing the warnings.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14770

Differential Revision: D13327629

Pulled By: suo

fbshipit-source-id: b9913f533d4530eb7c29146c39981ba7f72b6b68

Add output info when doing onnxGetBackendCompatibility (#14784)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14784

TSIA. To give more complete info to `onnxGetBackendCompatibility`.

Reviewed By: bertmaher, rdzhabarov

Differential Revision: D13331989

fbshipit-source-id: 1064b93f7f474788f736e6f0c893dae915c6fb99

Don't DCE PythonOp

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14773

Reviewed By: eellison

Differential Revision: D13327673

Pulled By: suo

fbshipit-source-id: 236db3407c7eacac470530836e3d4d0dc323110c

Improvements for symbolic AD (#14758)

Summary:
**Review only the last commit.**

This commit adds a few optimizations to AD, that let us dramatically
reduce the number of sizes we capture from forward.

We now:
- collapse chains of SumToSize
- avoid capturing sizes of tensors that are captured anyway
- more aggressively DCE the reverse code
- run CSE on the primal code to deduplicate `aten::size` calls

cc zou3519 zdevito
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14758

Differential Revision: D13324440

Pulled By: zou3519

fbshipit-source-id: 45ccbc13605adcef2b461840c6089d3200000c72

Revert D13289919: [pytorch][PR] [DataLoader] Refactor dataloader.py

Differential Revision:
D13289919

Original commit changeset: d701bc7bb48f

fbshipit-source-id: c350c491fefa98a0a7c0cf22cb832e78aeb15c3d

Delete defunct files from torch/csrc/distributed (#14785)

Summary:
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14785

Differential Revision: D13333066

Pulled By: ezyang

fbshipit-source-id: e7937b4e8e12409b0fa964c34f995f7861ca95ff

support conv transpose in script

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14775

Differential Revision: D13330491

Pulled By: eellison

fbshipit-source-id: 432b327d6a33517ff53ea33c9f64700e81432332

Making dist.get_default_group private for PT1 release (#14767)

Summary:
When I wrote the frontend API, it is designed on not letting users use the default_group directly on any functions. It should really be private.

All collectives are supposed to either use group.WORLD, or anything that comes out of new_group. That was the initial design.

We need to make a TODO on removing group.WORLD one day. It exists for backward compatibility reasons and adds lots of complexity.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14767

Reviewed By: pietern

Differential Revision: D13330655

Pulled By: teng-li

fbshipit-source-id: ace107e1c3a9b3910a300b22815a9e8096fafb1c

Make checkpoint_sequential work with multiple arguments (#14278)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14278

In this commit, we make checkpoint_sequential work for models with multiple tensor inputs. Previously, it only processed the first tensor and ignored the rest.

We introduce a new test in test/test_utils.py that replicates the issue referenced in this [GitHub issue](https://github.com/pytorch/pytorch/issues/11093), and we make sure that the test passes by changing the behavior of checkpoint_sequential to process all input tensors.

Reviewed By: ezyang

Differential Revision: D13144672

fbshipit-source-id: 24f58233a65a0f5b80b89c8d8cbced6f814004f7

update of fbcode/onnx to 42804705bdbf179d1a98394008417e1392013547 (#14777)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14777

Previous import was 6b34743d2e361bbc0acb29dd73536478cb92562e

Included changes:
- **[4280470](https://github.com/onnx/onnx/commit/4280470)**: Changes done internally at Facebook (#1668) <Lu Fang>
- **[f85221f](https://github.com/onnx/onnx/commit/f85221f)**: Fuse MatMul and Add into Gemm (#1542) <vloncar>
- **[022230e](https://github.com/onnx/onnx/commit/022230e)**: Replace np.long by np.int64 (#1664) <G. Ramalingam>
- **[0ab3c95](https://github.com/onnx/onnx/commit/0ab3c95)**: Infer shape from data in Constant nodes (#1667) <Shinichiro Hamaji>

Reviewed By: bddppq

Differential Revision: D13330082

fbshipit-source-id: 13cf328626cf872d0983bbd2154d95c45da70f1c

Enable testing on Loss modules (#14778)

Summary:
This PR adds `None` buffers as parameters (similarly to #14715). It also cleans up a bunch of the `test_jit.py` tests that should be covered by `common_nn.py` and brings in `criterion_tests` to test loss functions.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14778

Differential Revision: D13330849

Pulled By: driazati

fbshipit-source-id: 924cc4cf94e0dcd11e811a55222fd2ebc42a9e76

Add tests for dropout/batchnorm train/eval, remove training constants (#14780)

Summary:
This PR:

1. add tests for batchnorm/dropout for train/eval parameter mutatino
2. remove training constants from all our standard library
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14780

Differential Revision: D13331578

Pulled By: wanchaol

fbshipit-source-id: d92ca3ce38cc2888688d50fe015e3e22539a20a5

Split LegacyDeviceTypeInit from LegacyTypeDispatch. (#14723)

Summary:
The goal here is to have LegacyTHDispatch call into this as well, so LegacyTypeDispatch and LegacyTHDispatch don't have cross dependencies.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14723

Reviewed By: ezyang

Differential Revision: D13314017

Pulled By: gchanan

fbshipit-source-id: 8761cb4af2b2269d2e755203e073bfdba535b8c0

don't allow cse to clean up nondeterministic nodes

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14776

Differential Revision: D13330229

Pulled By: suo

fbshipit-source-id: 6bc88811e1889949f0f079cffccd8cd4270584cc

Reenable all forward-pass fusions that worked before the AD fix (#14558)

Summary:
Dealing with so many `aten::size` calls (in particular calls on elements computed inside fusion groups) requires us to do some extra graph processing in the fuser (to compute the sizes by explicit broadcasts, instead of writing the intermediate tensors only to check their size). This restores the forward expects of LSTM and MiLSTM to a single big kernel. Unfortunately the backward is much harder, because as long as we can't prove that the reductions are unnecessary (or if we can't distribute them over the op), we will not be able to fuse them.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14558

Differential Revision: D13321748

Pulled By: zou3519

fbshipit-source-id: c04fc2f70d106d2bfb56206b5aec517a93b79d1f

BatchNorm support not tracking stats

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14764

Differential Revision: D13325800

Pulled By: driazati

fbshipit-source-id: a3e4773dc31b83565e7a4de33614d6efd4a12de9

Minor doc change in c10/Device.h (#14762)

Summary:
Make sure it's a valid regex.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14762

Reviewed By: zrphercule

Differential Revision: D13326108

Pulled By: houseroad

fbshipit-source-id: fdcae2d5d42774c4071651b7477f08047d385dfa

Introduce LegacyTHDispatcher for dispatching to TH functions. (#14754)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14754

This isn't hooked up to anything yet, this is just putting the skeleton in place.
The idea here is that the functions generated via Declarations.cwrap and nn.yaml are not actually operators, they are implementation details of operators, and thus don't need to participate in VariableType, JIT dispatch generation.

So, we will split these functions out from the usual Type/operator hierarchy; for now the dispatch will be done by a Type-like class called LegacyTHDispatcher. Once this is done this probably means we can collapse Type to be backend-specific, not Type/ScalarType specific, because all the ScalarType specific code will live in the LegacyTHDispatcher.

Reviewed By: ezyang

Differential Revision: D13321605

fbshipit-source-id: 25d1bbc9827a42d6ab5d69aabbad3eac72bf364c

disable batch mm if we have mutable ops (#14771)

Summary:
Just to be safe, disable batch mm for mutable ops. We don't lose much for doing this, and we can go back at a calmer time to re-enable.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14771

Reviewed By: eellison

Differential Revision: D13327641

Pulled By: suo

fbshipit-source-id: 96611e21ed3cb8492a2cd040f7d33fb58c52bd5e

Replace at::Half non-vectorized conversions with implementations from FP16 (#14411)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14411
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14579

Folded the fp16 codes into c10.

Reviewed By: ezyang

Differential Revision: D13206450

fbshipit-source-id: 472208dd230dc49d33935622ff3286b17eeb0894

Use .to to convert new tensors in new_tensor (#14097)

Summary:
This would solve the tracing problems of #13969.
Fixes: #14732

I would appreciate if this got good scrutiny before applied.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14097

Differential Revision: D13323181

Pulled By: ezyang

fbshipit-source-id: dcd104b497c0bfddb751923c6166a3824b7a3702

Export generator constructor (#14041)

Summary:
Missed a spot :)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14041

Reviewed By: ezyang

Differential Revision: D13283803

Pulled By: ebetica

fbshipit-source-id: 482e245f57b0cea6ca3886355ea3ae487d024d4b

c10d doesn't work with torch namespace (#14042)

Summary:
If both `Utils.hpp` and the `torch` namespace is included in the same file, the compiler won't know which fmap to use. I believe this is because of ADL. This change fixes that issue for me.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14042

Reviewed By: ezyang

Differential Revision: D13283810

Pulled By: ebetica

fbshipit-source-id: b68233336518230ba730e83ddac1226a66896533

Add resnet test, convert more modules (#14437)

Summary:
This PR add resnet to test_jit and convert more nn modules, stacked on #14533 and #14715
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14437

Differential Revision: D13325871

Pulled By: wanchaol

fbshipit-source-id: 6c94a988b36794a373af6541c0c262a07291f7b1

Add missing test skip

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14763

Differential Revision: D13325350

Pulled By: driazati

fbshipit-source-id: 4d64a7616b227983c2fc2748c5fbecd1bcbff832

Rename _local_scalar to item() (#13676)

Summary:
Make `at::_local_scalar` more "official" by renaming it to `item()`.

gchanan
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13676

Differential Revision: D13003020

Pulled By: goldsborough

fbshipit-source-id: 0ac25f5237fb81a1576304a0a02f840ff44168a4

Remove use of hipify_caffe2, in favor of file path test. (#14757)

Summary:
This is towards unifying build_pytorch_amd.py and build_caffe2_amd.py
scripts. There is only one use of hipify_caffe2 left, which is just
to control which files actually get HIPified.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14757

Differential Revision: D13323486

Pulled By: ezyang

fbshipit-source-id: 958cd91be32dfc3c0a9ba9eda507adb5937aebcd

Add inplace FeedTensor for python frontend (#14512)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14512

att

Reviewed By: dzhulgakov

Differential Revision: D13243278

fbshipit-source-id: 78af417d0fcd9b9791ee839d62095903e49205cb

Loss (#14720)

Summary:
Adding Loss modules to script. Some of the modules have an optional tensor parameter. I will wait until wanchao's diff to support optional tensors is landed before landing this.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14720

Differential Revision: D13317990

Pulled By: eellison

fbshipit-source-id: 535925bdf126d28d9e7d64077b83ebd836a5beba

Add new reduction mode in kl_div (#14457)

Summary:
Fixes #6622 .
We used to average over all elements for kl divergence, which is not aligned with its math definition.
This PR corrects the default reduction behavior of KL divergence that it now naverages over batch dimension.

- In KL, default behavior `reduction=mean` averages over batch dimension. While for most other loss functions, `reduction=mean` averages over all elements.
- We used to support scalar tensor as well. For BC purpose, we still support it, no reduction is performed on scalar tensor.
- Added a new reduction mode called `batchmean` which has the correct behavior for KL. Add a warning to make `batchmean` as default for KL instead of `mean` in next major release.
- [deprecated]I chose to not add a new reduction option, since "mean over batch dimension" is kinda special, and it only makes sense in few cases like KL. We don't want to explain why there's a option "batchmean" but it's not applicable for all other functions. I'm open to discussion on this one, as I cannot think of a perfect solution for this.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14457

Differential Revision: D13236016

Pulled By: ailzhang

fbshipit-source-id: 905cc7b3bfc35a11d7cf098b1ebc382170a087a7

Implements Gather operator for arbitrary axis, sharing the code with BatchGather. (#13756)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13756

This implements general Gather operator for arbitrary axis, sharing the code with BatchGather.
- CPU gather & batch gather logic is now shared through caffe2::gather_helper, for any axis.
- Shared CUDA kernel moved to gather_op.cuh, for any axis.
- Gradients of axis > 0 delegate to BatchGatherGradientOp which now has axis argument.
- BatchGatherOp doc strings updated to have correct rank (q + (r -1)) and output.
- Added tests for axis == 2.

GatherOp supports index wrapping for axis == 0 by default, which was earlier for ONNX.
This diff also extends it to work in Cuda kernel. Added "wrap_indices" argument which specifies
wheather this wrapping should be done; set it to true if you'd like wrapping for any axis.

TBD: Update gradients to support negative indices (separate diff).
TBD: Once we have operator versioning, we'd like to update GatherOp to NOT support axis 0 wrapping
by default, but rather do it only if wrap_indices is set.

Reviewed By: dzhulgakov

Differential Revision: D12983815

fbshipit-source-id: 8add9d67b47fe8c5ba7a335f581ca0530b205cd7

Refactor dataloader.py (#14668)

Summary:
As I am working on tasks in https://github.com/pytorch/pytorch/issues/13023, I realized how unreadable the code is because all functions to be run in multiprocessing must be at top global level. Adding more functionalities to `dataloader.py` will only make things worse.

So in this PR, I refactor `dataloader.py` and move much of it into `data._utils`. E.g., the `_worker_loop` and related methods are now in `data._utils.worker`, signal handling code in `data._utils.signal_handling`, collating code in `data._utils.collate`, etc. This split, IMHO, makes code much clearer. I will base my future changes to DataLoader on top of this.

No functionality is changed, except that I added `torch._six.queue`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14668

Reviewed By: soumith

Differential Revision: D13289919

Pulled By: ailzhang

fbshipit-source-id: d701bc7bb48f5dd7b163b5be941a9d27eb277a4c

Back out "Move TensorOptions, DefaultTensorOptions and OptionsGuard to c10" (#14745)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14745

Original commit changeset: c62e7f9b0255

Reviewed By: suo

Differential Revision: D13318594

fbshipit-source-id: 4d7dc35ca01b627accc3ee512bfcd6f2e805a533

Back out "Fix include paths for TensorOptions, DefaultTensorOptions, OptionsGuard" (#14744)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14744

Original commit changeset: d236d5351ecf

Reviewed By: suo

Differential Revision: D13318596

fbshipit-source-id: 55f1e9472d05fb5a9c47dc82c32e9a66b5e4308c

Disable randn_like fusion in the JIT (#14752)

Summary:
Fixes #14674. We won't have time for a proper fix before the release, so at least disable fusion of nodes that trigger incorrect behavior.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14752

Differential Revision: D13320407

Pulled By: zou3519

fbshipit-source-id: 2400f7c2cd332b957c248e755fdb0dadee68da5d

fix import failure in hub test (#14742)

Summary:
Fix #14610

I can repro the test failure following the steps provided, and this fixes the issue for me. Seems the timing of inserting has to happen after the downloading.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14742

Differential Revision: D13318533

Pulled By: ailzhang

fbshipit-source-id: b9207b4572d5a9443e516d9a84632e3d7b68e477

Revert D13304654: [pytorch][PR] Introduce LegacyTHDispatcher for dispatching to TH functions.

Differential Revision:
D13304654

Original commit changeset: cfe3e1a28adc

fbshipit-source-id: 06669d3c88f83e1d959e2c266fd608316539d42a

Introduce LegacyTHDispatcher for dispatching to TH functions. (#14708)

Summary:
This isn't hooked up to anything yet, this is just putting the skeleton in place.
The idea here is that the functions generated via Declarations.cwrap and nn.yaml are not actually operators, they are implementation details of operators, and thus don't need to participate in VariableType, JIT dispatch generation.

So, we will split these functions out from the usual Type/operator hierarchy; for now the dispatch will be done by a Type-like class called LegacyTHDispatcher. Once this is done this probably means we can collapse Type to be backend-specific, not Type/ScalarType specific, because all the ScalarType specific code will live in the LegacyTHDispatcher.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14708

Reviewed By: ezyang

Differential Revision: D13304654

Pulled By: gchanan

fbshipit-source-id: cfe3e1a28adcc355f67fe143495ee7e5c5118606

add .code property to ScriptModule (#14735)

Summary:
simple change to allow `print(foo.code)` to give a pretty-printed description of all the methods on a module.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14735

Differential Revision: D13317619

Pulled By: zdevito

fbshipit-source-id: dc7f7ba12ba070f2dfccf362995c2a9e0e573cb7

Fix clamp when min/max are both None (#14716)

Summary:
Before this PR, tensor.clamp() would return an empty tensor if min and
max were not specified. This is a regression from 0.4.1, which would
throw an error. This PR restores that error message.

Fixes #14470
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14716

Differential Revision: D13311031

Pulled By: zou3519

fbshipit-source-id: 87894db582d5749eaccfc22ba06aac4e10983880

Restore device in cpp API (#14711)

Summary:
This is a stack PR based on https://github.com/pytorch/pytorch/pull/14454.

It enables the restoring the storage to appropriate device.

~~[TODO]: add/modify appropriate tests~~ Done
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14711

Reviewed By: dzhulgakov

Differential Revision: D13315746

Pulled By: houseroad

fbshipit-source-id: fe6f24a45c35e88fd1a2eebc09950d4430fac185

move structs to header file (#14728)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14728

Move IndexBlob,Index to header file so it can reused.

Differential Revision: D13315898

fbshipit-source-id: 34432c9b8fa08af3d3387f32a940d35b02a59760

improve the restore device test, and relax the assertion (#14734)

Summary:
Only compare the device index if device has it.

Test the tensor restore with some computation.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14734

Reviewed By: dzhulgakov

Differential Revision: D13317949

Pulled By: houseroad

fbshipit-source-id: 26b2f2912a9bbc3b660a62283fb403ddab437e49

Reduce broadcasted inputs in derivative code (#14485)

Summary:
Previously symbolic AD formulas assumed that no broadcasting happened,
and would return gradients of incorrect shapes (possibly leading to
silent errors later).

Fixes a few bugs (known and unknown):
- #11736
- ArgumentSpec didn't compute the input types correctly [(it didn't advance the offset for non-tensor args)](https://github.com/pytorch/pytorch/pull/14485/files#diff-4fd3157a056596aefb8cdf41022a208bR153)
- Symbolic AD could suffer from use after free (dangling pointers in grad map), because [`EliminateDeadCode` could have removed nodes](https://github.com/pytorch/pytorch/pull/14485/files#diff-25d33ad1ed6855684dec79d927ca6142L781) that referenced gradients of certain values.
- Undefined behavior in `aten::size`

During my tests I've also found a few new problems, and I have opened issues for them:
- FusionGroup seems to think that cat nodes broadcast their inputs (#14483)
- `prim::ConstantChunk` derivative formula doesn't handle undefined inputs (#14484)

This patch unfortunately deoptimizes some of our code (Fusion doesn't happen past chunk nodes, and outputs more tensors only because we have to get their size). I know how to fix those issues, but wanted to fix this terrible bug quickly.

cc zou3519 zdevito ngimel
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14485

Reviewed By: eellison

Differential Revision: D13312888

Pulled By: suo

fbshipit-source-id: ad46bfb4d0a306ad9451002f8270f7a790f72d58

interpolate (#14123)

Summary:
Add support for interpolate and upsampling in weak_script mode.

Because the function parameters are overloaded, i had to add it as a builtin op. For interpolate:
size can be ?int | int[]?, and scale_factor can be ?float | float[]?. Every combination of the two parameters needs to be supported.

The same logic applies for upsample_nearest, upsample_bilinear, and upsample.

There are a few fixes that I came to along the way.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14123

Differential Revision: D13278923

Pulled By: eellison

fbshipit-source-id: e59729034369be4ce4b747291a3d1c74e135b869

Add Pooling modules to Script (#14527)

Summary:
Depends on #14584
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14527

Differential Revision: D13270773

Pulled By: driazati

fbshipit-source-id: e4acd43ccbce0f4b62d41c30ce8d5c721171e19a

Add fractional_max_pool2d to standard lib

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14591

Differential Revision: D13270755

Pulled By: driazati

fbshipit-source-id: 138a60256795f5ef8d236c75be2cfd929059b98f