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

A trivial error message updates on `at::Tensor _convolution` (#15830)

Summary:
I fixed an grammatical error on this function previously, but I also realized that its content was also wrong. A weight tensors of a convolutional layer should be at least 3 dimensional, not 2.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15830

Differential Revision: D13597968

Pulled By: soumith

fbshipit-source-id: 72a75106e88945c68d6462828b149441cfb5acde

Enable torch static build on Windows

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15769

Reviewed By: yf225, pjh5

Differential Revision: D13597845

Pulled By: orionr

fbshipit-source-id: 99640e22974990ae570a4795ce07274c4447cb01

Fix sum_to behavior with zero dimensions (#15796)

Summary:
Fixes #15223.

This fixes an autograd bug where backprop either fails or produces
gradients of incorrect sizes when tensors with zero-sized dimensions are
involved.

Previously, we were reducing along dimensions that had size greater than 1
when summing to a size in autograd. This is incorrect because we should also reduce
along dimensions with size 0 to produce a tensor of size 1 in that
dimension that then gets viewed to the correct shape.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15796

Differential Revision: D13593199

Pulled By: zou3519

fbshipit-source-id: 2e2acac34943a9b7fabadc10c9efd4f66db298fd

Cache workspace size in the BenchmarkCache. (#15742)

Summary:
Cache the workspace size information for MIOpen for a given configuration as opposed to inquiring it every time. This reduces overhead significantly as inquiring the workspace size forces a full read of the performance database in MIOpen and this database has grown significantly in recent releases. This caching gets us back to ideal performance.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15742

Differential Revision: D13598932

Pulled By: bddppq

fbshipit-source-id: 4e65d247b71dec828293cf0562aac3fbd4fad83a

Refactors shape logic out of code generation, fixes possible segfault (#15750)

Summary:
This PR:

- Removes shape logic from the code generator, which was previously relied on to return chunk and concat information
- Copies the logic to detect if a kernel has a rand_like node to the executor, making its pass independent of the code generator
- Fixes a possible segfault where references to a vector still being modified were relied upon

The actual shape logic is unchanged.

The possible segfault is in the handling of the former "flat_inputs" in codegen.cpp. This vector holds pairs, and the second element of these pairs is a reference. In some cases these would be references to items in the vector chunk_desc, which could be added to later, possibly invalidating any references to items in it. I hit a similar segfault in testing when naively making parallel code for "flat_outputs."

I'm submitting this small PR because it's separable, self-contained, has a fix, and I am trying to actively get away from large PRs to encourage more stability and incremental change in the fuser.

ngimel zou3519
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15750

Differential Revision: D13597451

Pulled By: zou3519

fbshipit-source-id: 0d48b365779b42849b044ba0286258aacc7b0332

Use parallel thrust execution policy on ROCm (#15481)

Summary:
The Thrust shipped with ROCm is recent enough to support this API. Minimize divergence between CUDA/ROCm by changing idef guards.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15481

Differential Revision: D13598739

Pulled By: bddppq

fbshipit-source-id: 20d0a7e3887a4050eea65033161561af47411de1

Use correct workspace alloc call in MIOpen conv operator (#15712)

Summary:
This PR contains changes for:
1. Using memory alloc from HIPContext while allocating workspace for MIOpen conv and transpose_conv operators rather than direct HIP mem alloc
2. Minor cleanup and removing an unnecessary sync call from MIOpen conv op

Differential Revision: D13598894

Pulled By: bddppq

fbshipit-source-id: 44886161abdf91cd29c7c93b3e23620e1b09c7c9

Tensor method rename dims()->sizes() - 2/2

Summary: Codemod generated with clangr shard mode, 25 files per diff,

Reviewed By: smessmer

Differential Revision: D13581787

fbshipit-source-id: b04c6aa87fea3a10b522a71fccc1fcfb76a2c212

Remove caffe2::ShareData (#15418)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15418

Previously we are using Resize + ShareData.
Instead, we'll create a function on Tensor that clones itself with same storage.

Suppose we want `t` to `ShareData` with `t0`, Previous:
```
Tensor t(dims, CPU);
t.Resize(t0.sizes());
t.ShareData(t0);
```
Now:
```
Tensor t = t0.Alias();
```

Reviewed By: dzhulgakov

Differential Revision: D13507609

fbshipit-source-id: 6e4275d02f4c3356cbce91127f1b01111dc86b9f

Move isnan to C++ (#15722)

Summary:
Wanted to use `Tensor.isnan` in C++, figured it'd be nice to have, so I made it into a tiny native function.

gchanan ezyang apaszke
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15722

Differential Revision: D13591315

Pulled By: goldsborough

fbshipit-source-id: a78bd22101fde87a0257f759b9bfcf3b4208f5fa

use all_weights instead of _parameters in _flat_weights in rnn (#15766)

Summary:
Fixes #15749
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15766

Differential Revision: D13592320

Pulled By: soumith

fbshipit-source-id: 6c3805f576c3df5a2da8bef1e4305eda379718df

Use CUDAGuard when serializing CUDA Tensors (#15807)

Summary:
Fixes #15308. Before this change, `torch.save` and `torch.load` would
initialize the CUDA context on GPU 0 if it hadn't been initialized
already, even if the serialized tensors are only on GPU 1.

This PR fixes that bug by using CUDAGuard in the storage serialization
path.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15807

Differential Revision: D13593201

Pulled By: zou3519

fbshipit-source-id: 4addc91ea5a5278d56a03f3d422577ee39e99897

Stop leaving garbage files after running test_jit.py

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15404

Differential Revision: D13548316

Pulled By: zou3519

fbshipit-source-id: fe8731d8add59777781d34d9c3f3314f11467b23

Add support for batch_norm fusion to the JIT (#15146)

Summary:
We don't support reductions yet, but simply decomposing batch_norm
into a kernel that computes the stats, and the fusing everything else
with ReLU and following pointwise ops provides nice speedups.

Note that this is only limited to inference mode for now, because we
don't support convolutions and batch norm in AD, so the fuser isn't
applied to those parts.

This commit gives us a 7% end-to-end speedup for ResNet50 with batch size 32. Note that this only applies to inference mode at the moment due to lack of AD support for CNN operations (I'll be adding that soon), and not to the standard `torchvision` models, because they use in-place ops which aren't supported by the fuser (we need a way of proving that de-inplacing them is safe).

cc zou3519 zdevito mruberry ngimel
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15146

Differential Revision: D13548303

Pulled By: zou3519

fbshipit-source-id: a2e2e5abc383f637fae19bd1b423f20c2cbc056a

Support communicating with C2 protobuf in Onnxifi flow (#15472)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15472

Create a path to pass serialized C2 protobuf instead of ONNX during ONNXIFI flow

Reviewed By: houseroad

Differential Revision: D13536603

fbshipit-source-id: 7d016474f4beedbda480ed2e2c0004af7868aafe

Add count_include_pad arg for AveragePoolOp on GPU (#15787)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15787

Add count_include_pad arg for AveragePoolOp on GPU

Reviewed By: houseroad

Differential Revision: D13589185

fbshipit-source-id: 235a84cfcd2033ee796c13e338fc3d03e832b5b1

Move Stream.query() implementation down to C++ (#15737)

Summary:
See #15682

Pushing up this small PR to check if I am doing the right thing. If correct, more will follow for other Stream APIs. Questions will be added inline.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15737

Differential Revision: D13581400

Pulled By: mrshenli

fbshipit-source-id: 24afed7847b89b62f0692c79a101ec7ff9d9ee4d

A trivial error in the error message of `at::Tensor _convolution` fixed (#15772)

Summary:
A trivial grammatical error fix.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15772

Differential Revision: D13592279

Pulled By: zou3519

fbshipit-source-id: 14f60c61747a3893cd0e4c860f7b4c4c4ba28c28

clean up D13579188 (#15759)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15759

Some flags have too long names. And some other few minor clean ups.

Reviewed By: jianyuh

Differential Revision: D13587353

fbshipit-source-id: f8aee7f167505644f5d8f80fe2eed70201ef1e54

Add support for exporting onnx split (#15092)

Summary:
* With the update of split output to dynamic list it breaks the export to onnx.
Now split ir becomes two ops: 1. Dynamic[] <= Split(), and 2. out1, out2, out3
<= Prim::ListUnpack. In this fix these two consecutive ops get fused when being
exported to onnx.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15092

Reviewed By: dzhulgakov

Differential Revision: D13583832

Pulled By: houseroad

fbshipit-source-id: 3eb18c871e750921ad6d5cc179254bee9bcf4c99

simplify conv dnnlowp ops by not allowing fp32 in/out (#15758)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15758

DNNLOWP Conv operators became very complex due to many options. This diff simplifies them by not allowing fp32 in/out. This is OK for Conv operators because Conv operators are usually used in deep networks where quantizing and dequantizing using separate operators is not much overhead.

Reviewed By: csummersea

Differential Revision: D13587341

fbshipit-source-id: e88c919dae79d1c5b7d787ea539edf5bcb064afc

Enable conv+add fusion, same as conv+sum (#15268)

Summary:
Enable conv+add fusion, same as conv+sum

Caution: only element-wise add is supported on IDEEP without scalar
broadcast. Otherwise, the fusion is illegal.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15268

Differential Revision: D13577375

Pulled By: yinghai

fbshipit-source-id: 92c9c4b667c5ca5f7a262a5bffaa8aa68eeff3bd

Allow List arguments to Python Ops (#15721)

Summary:
Adds `List` to eval environment for type lines and allows `List` to be used on PythonOps (follows the same style as the `Tuple` code), fixes #15661
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15721

Differential Revision: D13578540

Pulled By: driazati

fbshipit-source-id: fce54dc3c0931d8b017b2e3483f0ac53826dda94

Bump CircleCI docker version to 278 (#15795)

Summary:
Just changing the version number doesn't seem to work. I needed to also fix macos brew parallel conflict

should this merge together with https://github.com/pytorch/ossci-job-dsl/pull/36 ?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15795

Differential Revision: D13591839

Pulled By: yf225

fbshipit-source-id: 6b2a90943e63c8dcc4b6d9159eb54f1b5974c9ac

Fix C++ Frontend example in frontend.html (#15717)

Summary:
The small end-to-end example in https://pytorch.org/cppdocs/frontend.html is a little outdated and needs fixes.

ezyang soumith
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15717

Differential Revision: D13591306

Pulled By: goldsborough

fbshipit-source-id: 3334d68c7f77cf094b66ec2b2f396c4c65bb0d72

Fix restructured text issue in tensor_basics.rst (#15701)

Summary:
Fix submitted by huntzhan in https://github.com/pytorch/cppdocs/pull/4. The source is in this repo so the patch has to be applied here.

soumith ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15701

Differential Revision: D13591302

Pulled By: goldsborough

fbshipit-source-id: 796957696fd560a9c5fb42265d7b2d018abaebe3

Fallback to CPU concat op to handle TensorCPU inputs (#15263)

Summary:
Fallback to CPU concat op to handle TensorCPU inputs
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15263

Differential Revision: D13587030

Pulled By: yinghai

fbshipit-source-id: 010a8579d61c3beb8556eb92493a552b2ab0030c

fix conv unit test for groupwise quantization and pre-packing (#15761)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15761

As title says.

Reviewed By: csummersea

Differential Revision: D13587727

fbshipit-source-id: f0631b8cbb89d65a1d952bc25b463de23de93bec

Add is_floating_point to docs (#15704)

Summary:
Fixes #15700 .

Changelog:

- Expose torch.*.is_floating_point to docs

Differential Revision: D13580734

Pulled By: zou3519

fbshipit-source-id: 76edb4af666c08237091a2cebf53d9ba5e6c8909

Pool prim::None nodes (#15745)

Summary:
Make the constant pooling pass pool prim::None nodes
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15745

Differential Revision: D13583518

Pulled By: eellison

fbshipit-source-id: 7f8aa70522515805ab0991c6db3d96b5a96cdede

Replace some malloc+memset pairs with calloc.

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15765

Differential Revision: D13588723

Pulled By: resistor

fbshipit-source-id: 47d35dc608847a5b173cfcf2aaa2a77359e56722

Removes print statements from test_torch.py (#15747)

Summary:
These print statements do not affect the test, and tests (generally) shouldn't print.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15747

Differential Revision: D13587289

Pulled By: soumith

fbshipit-source-id: c758793c9e35faf02bacba6c7c6d072f7c40453f

Fix several DeprecationWarning: invalid escape sequence (#15733)

Summary:
Hello,

This is a little patch to fix `DeprecationWarning: invalid escape sequence`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15733

Differential Revision: D13587291

Pulled By: soumith

fbshipit-source-id: ce68db2de92ca7eaa42f78ca5ae6fbc1d4d90e05

caffe2_benchmark msvc build fix (#15619)

Summary:
Fixing error in caffe2_benchmark binary

```
2018-12-29T14:09:59.7867995Z d:\a\1\s\caffe2_builders\v141\pytorch\binaries\benchmark_helper.h(90): error C2678: binary '|=': no operator found which takes a left-hand operand of type 'std::_Iosb<int>::_Openmode' (or there is no acceptable conversion) (compiling source file D:\a\1\s\caffe2_builders\v141\pytorch\binaries\benchmark_helper.cc) [D:\a\1\s\caffe2_builders\v141\pytorch\build\Release\binaries\caffe2_benchmark.vcxproj]
2018-12-29T14:09:59.7868252Z d:\a\1\s\caffe2_builders\v141\pytorch\binaries\benchmark_helper.h(92): error C2678: binary '|=': no operator found which takes a left-hand operand of type 'std::_Iosb<int>::_Openmode' (or there is no acceptable conversion) (compiling source file D:\a\1\s\caffe2_builders\v141\pytorch\binaries\benchmark_helper.cc) [D:\a\1\s\caffe2_builders\v141\pytorch\build\Release\binaries\caffe2_benchmark.vcxproj]
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15619

Differential Revision: D13580195

Pulled By: soumith

fbshipit-source-id: b0a4479cd5f7555801b1977aeee96b6433293da7

Adding a hook (wrapper) for non-std stream reader in PyTorchStreamReader (#15551)

Summary:
To implement a stream is very annoying, since it is closely defined with the underlying storage streambuffer.

So in this PR, we add ReadAdapterInterface and PyTorchStreamReader will use it. We implement IStreamAdapter as a wrapper of std::istream. And keep the user interface unchanged.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15551

Reviewed By: zrphercule

Differential Revision: D13568907

Pulled By: houseroad

fbshipit-source-id: 93708cb801248a6c101f35cb14d1631029365c3c

support 0 size in any of the tensor dimensions in mkldnn (#15295)

Summary:
support 0 size in any of the tensor dimensions in mkldnn
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15295

Differential Revision: D13573747

Pulled By: yinghai

fbshipit-source-id: 5bf7a0b9e2567e80f44981a7823be5407fc94e53

Port replication_pad2d and replication_pad3d to ATen (#15538)

Summary:
port replication padding 2D and 3D from legacy TH API implementation
to ATen implementation.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15538

Differential Revision: D13547567

Pulled By: lhuang04

fbshipit-source-id: decfe100d9edfdcfb62f39ee23f37b6cae0d461f

Fix different types in rsub caused bug (#15707)

Summary:
Before this pr, rsub did not convert two elements into the same dtype, therefore "1 - x" may export to an onnx model that two elements of rsub having different dtype.
By adding this symbolic patch this bug should be fixed.
Related test cases also created.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15707

Differential Revision: D13583042

Pulled By: zrphercule

fbshipit-source-id: 3a2de47a1a8d1ded1a0adfb911adbe6ac729cdef

Tensor method rename dims()->sizes() - 1/2

Summary: Codemod generated with clangr shard mode, 25 files per diff,

Reviewed By: BIT-silence

Differential Revision: D13581782

fbshipit-source-id: b16b4198e100617769d84aa599bf141117cfbe5b

update of fbcode/onnx to 8384c788939bc65463f9754b6a7a00b212b18ba1 (#15739)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15739

Previous import was 765f5ee823a67a866f4bd28a9860e81f3c811ce8

Included changes:
- **[8384c78](https://github.com/onnx/onnx/commit/8384c78)**: add constantofshape (#1582) <Rui Zhu>
- **[9afc06c](https://github.com/onnx/onnx/commit/9afc06c)**: Set symbol visibility to hidden for non-Windows (#1707) <Paul Jesse Hellemn>
- **[6f8a9f0](https://github.com/onnx/onnx/commit/6f8a9f0)**: Revert "Add NonMaxSupression operator (#1695)" (#1702) <Lu Fang>
- **[8b89544](https://github.com/onnx/onnx/commit/8b89544)**: Add NonMaxSupression operator (#1695) <Hector Li>
- **[0a7cc48](https://github.com/onnx/onnx/commit/0a7cc48)**: Add bfloat16 support. (#1699) <Dmitri Smirnov>
- **[da7c50c](https://github.com/onnx/onnx/commit/da7c50c)**: ONNX does not maintain versions for experimental ops (#1696) <Ke Zhang>
- **[0c8d857](https://github.com/onnx/onnx/commit/0c8d857)**: Correct type of value_info in Graph (#1694) <Maik Riechert>
- **[f612532](https://github.com/onnx/onnx/commit/f612532)**: Fix typos (#1686) <Eundoo Song>

Reviewed By: zrphercule

Differential Revision: D13581674

fbshipit-source-id: 8f8ee86a05a86fe99bf94509148c559ea3df1464

remove use of tmp_install

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14553

Differential Revision: D13583335

Pulled By: anderspapitto

fbshipit-source-id: 8711fead9eda877c1037a0bc59f91a3d2e01f3e0

Update CI credentials

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15736

Differential Revision: D13583174

Pulled By: yf225

fbshipit-source-id: 742470db10ef9df8f95e27626453b68ca90723e8

Temporarily disable all XXXlike operator tests in pytorch-onnx test (#15740)

Summary:
We are going to have some breaking changes in ConstantLike and related operators in onnx, therefore it is better to disable all related tests for these operators for now.
These operators are not currently supported by caffe2, and are not included in our most recently released onnx, therefore we do not need to worry about internal/external production breaking.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15740

Differential Revision: D13582528

Pulled By: zrphercule

fbshipit-source-id: 92a890c1dc2a833969af69edfea85331bb4d562f

Tensor construction codemod - 2/2 (#15600)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15600

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

Reviewed By: dzhulgakov

Differential Revision: D13542455

fbshipit-source-id: 8a3b15b0a1f81565f34e309114e1c3e1f7f65a3c

Print out operator suggestions for unknown builtin op (#15183)

Summary:
This improves the error message for "unknown builtin op" to suggest similarly named ops.

Currently it prints out all operators with a name within two edits.

Related issue: https://github.com/pytorch/pytorch/issues/13409
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15183

Differential Revision: D13578509

Pulled By: eellison

fbshipit-source-id: 5c73408eda1f7aa456f5bd28790c34df0c76aeca

Updating submodules

Reviewed By: yns88

fbshipit-source-id: b8be56b57d109dfef5980ea7255e2ab021da099e

Tensor construction codemod - 1/2 (#15598)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15598

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

Reviewed By: dzhulgakov

Differential Revision: D13542429

fbshipit-source-id: db1059c78e85724d9b4fdab70466cf329db68359

remove dependency to fp32 batch permutation op (#15723)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15723

As title says.

Reviewed By: jianyuh

Differential Revision: D13578604

fbshipit-source-id: 0da0ac31ae83c1e0daa9077e878feb4deffed6a3

Cudnn Handle Pool 3: At Wit's End (#15668)

Summary:
ezyang Here's a freshly rebased version of https://github.com/pytorch/pytorch/pull/15080 with the if statement that relieved the hangs that occasionally, nondeterministically, occurred on cudnnCreate on a particular windows build ([example w/debug statements](https://ci.pytorch.org/jenkins/job/pytorch-builds/job/pytorch-win-ws2016-cuda9-cudnn7-py3-test2/19238/console)) in https://github.com/pytorch/pytorch/pull/15280.

I'd like to run the CI over this several times before it's considered mergeable. Sometimes the windows hang doesn't manifest for 2 or 3 consecutive trials.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15668

Differential Revision: D13579291

Pulled By: soumith

fbshipit-source-id: 3972eb98bad6ece933ca5e67a10fc4bc2ed06068

Remove TH/THC link for cholesky_solve (#15691)

Summary:
Changelog:
- Remove TH/THC binding
- Port single matrix case to ATen
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15691

Differential Revision: D13579317

Pulled By: soumith

fbshipit-source-id: 63a55606c656396e777e8e6828acd2ef88ed1543

Modify torch.gesv error message (#15654)

Summary:
[doc](https://pytorch.org/docs/stable/torch.html#torch.gesv) uses `B` uppercase so error message should follow to avoid confusion.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15654

Differential Revision: D13571297

Pulled By: soumith

fbshipit-source-id: 0b4e7797eceff92618f808bbfa65d13c1dcc2da0

make conv_depthwise_dnnlowp_op_test faster (#15725)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15725

As title says.

Reviewed By: jianyuh

Differential Revision: D13579188

fbshipit-source-id: 382072c95929ccf9e189e2338e35b046c4a0650f

clarified language of doc for torch.mul (#15664)

Summary:
see issue #15636

Please note - I build the documents but the HTML is not updated with the edited content.
I did not also build the fork.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15664

Differential Revision: D13571310

Pulled By: soumith

fbshipit-source-id: d43be0f61705693d778cc12c13e86d6b06130ac7

disallow nbits_in_non_outlier == 0 in acc16 conv; option to fallback to acc32 (#15708)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15708

nbits_in_non_outlier == 0 doesn't make sense because it means everything is outlier and we can just use 32-bit accumulation.
Depending on architecture, break-even point between acc16 and acc32 can be different. Adding thresholds for falling back to acc32.

Reviewed By: jianyuh

Differential Revision: D13574832

fbshipit-source-id: b7a37aacbfdc7867e31838dafcdd5f7c2ac282af

Torch tensor (#15224)

Summary:
Support torch.tensor in script. Already been accepted, trying to reland
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15224

Differential Revision: D13466616

Pulled By: eellison

fbshipit-source-id: f7850da07b0eb11af98f255fc15bd3cf861f2a40

A quick fix for Stream operation errors on non-current device (#15689)

Summary:
see #15682

This is a quick fix by implementing the simpler solution as suggested by colesbury. As benchmark result shows, it slows down `Stream.query()` by ~20%, I would be happy to further pursue a more complex solution by implementing this in C++/ATen. But I would still vote for merge this quick fix first just to get rid of the bug sooner.

~Test TBA~ Added

FYI jeffreyksmithjr

now

```python
In [1]: def f():
   ...:     d0 = torch.device('cuda:0')
   ...:     d1 = torch.device('cuda:1')
   ...:     with torch.cuda.device(d0):
   ...:         s0 = torch.cuda.current_stream()
   ...:     with torch.cuda.device(d1):
   ...:         s1 = torch.cuda.current_stream()
   ...:     s0.query()
   ...:     s1.query()

In [4]: %timeit f()
38.1 µs ± 4.2 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

In [5]: %timeit f()
37.6 µs ± 2.7 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)
```

before

```python
In [4]: %timeit f()
28.5 µs ± 1.74 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

In [5]: %timeit f()
35.3 µs ± 2.91 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15689

Differential Revision: D13571697

Pulled By: mrshenli

fbshipit-source-id: 4fe697f91248c6419136d37bb5b7147e612e2f4c

Break up generated tests (#13992)

Summary:
This PR breaks up `TestJitGenerated` into 3 classes. This makes for
easier testing of specific groups (e.g. run all generated functional
tests without having to wait for the autograd tests)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13992

Differential Revision: D13076371

Pulled By: driazati

fbshipit-source-id: 1267af59be7d69feb690f5805fcd43fea58a7159

flake8 hook fix (#15693)

Summary:
This PR bypasses checking the user's configuration entirely and always use strict, since the CI considers it a hard failure if you can't pass flake8.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15693

Differential Revision: D13574889

Pulled By: suo

fbshipit-source-id: f5e1c5731cc49b6223b415317033c275bc7d4fec

Prevent VS2017 from emitting ambiguous symbol errors (#15697)

Summary:
These `std::forward` calls cause VS2017 to emit:

error C2872: 'std': ambiguous symbol

This fix prevents the ambiguity by specifying that `::std` is intended.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15697

Differential Revision: D13573483

Pulled By: goldsborough

fbshipit-source-id: 0439de3523a37a18df7af0cff4a1284a53833ddd

trace s_copy_ (#15690)

Summary:
s_copy_ was previously special-cased for out of place tracing.
This adds support for inplace tracing, which fixes tracing of
inception_v3

Fixes #15216
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15690

Differential Revision: D13572011

Pulled By: zdevito

fbshipit-source-id: 1d565dec039a4b8c59179254285e61d2517ef9a9

Add mkldnn conv double backward (#15686)

Summary:
Fixes #15353 .

Like cudnn conv implementation, mkldnn also falls back to the default `_convolution_double_backward` as double backward.

This bug wasn't caught by CI before because mkldnn is only used when input scalar type is float, but our tests are all using double as default.

Adding test for float inputs, but mkldnn seems to have imprecision issues similar to cudnn implementation, so here I only check if double backward exists instead of calling `gradgradcheck`. Please correct me if the precision should actually be checked.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15686

Differential Revision: D13571682

Pulled By: ailzhang

fbshipit-source-id: f1762439762370f276cfd59e8b8b8a4dee960a4b

Fix ONNX export of logical ops, including torch.ne, to have correct output datatype (#15677)

Summary:
This is the an updated version of the earlier PR https://github.com/pytorch/pytorch/pull/15185, since that one was closed.

Currently PyTorch ONNX exporter exports the logical ops (lt, gt, le, ge, eq, ne) with output type in corresponding ONNX ops as type tensor(uint8). But ONNX spec allows for only tensor(bool), which is why models that have these ops fail to load properly.

This issue is captured in #11339. Part of this issue, relating to the allowed input types, has been fixed in ONNX spec by houseroad. This PR fixes the other part pertaining to output type.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15677

Reviewed By: dzhulgakov

Differential Revision: D13568450

Pulled By: houseroad

fbshipit-source-id: a6afbea1afdb4edad8f8b1bc492f50b14e5f2fce

Port legacy reflection_pad1d to ATen (#15480)

Summary:
1. Avoided using `THCDeviceTensor` by re-calculating the mapping from cuda (blockIdx, threadIdx) to input/output tensor index.
2. Changed Camelcase naming to underscore naming.

Profiling:

Legacy:

```bash
$py.test test/test_nn.py -k ReflectionPad1d -v -s
....
=========== 2 passed, 1258 deselected, 800 warnings in 4.35 seconds ============
```

Now:

```bash
$py.test test/test_nn.py -k ReflectionPad1d -v -s
...
=========== 2 passed, 1258 deselected, 800 warnings in 4.03 seconds ============
```

I have two questions about the code. Any insights are appreciated. gchanan zou3519

1. I can verify that [this magic](https://github.com/pytorch/pytorch/blob/master/aten/src/THCUNN/TemporalReflectionPadding.cu#L32-L36) correctly maps output index to input index in different cases. But, I have no idea about how did you come up with this algorithm that merges three categories (in left padding, in original input, in right padding) into a single statement?

2. Why do we need [get contiguous](https://github.com/pytorch/pytorch/blob/master/aten/src/THNN/generic/TemporalReflectionPadding.c#L80) tensors when calculating forward and backward propagation?

Reflection_pad2d porting will come in the next PR.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15480

Differential Revision: D13544924

Pulled By: mrshenli

fbshipit-source-id: 182045434f210032a82cab721a190da0cd781fbf

bug fix in 3d group conv (#15625)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15625

3D group conv (both NCHW and NHWC layout) was not correct.
Added group=2 in test_1d_convolution and test_3d_convolution in conv_test

Reviewed By: protonu

Differential Revision: D13562099

fbshipit-source-id: 586e8a7574a2764f2a3b559db6c2415b3ab90453

Port torch.arange to aten and parallelize on CPU.

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15667

Differential Revision: D13566631

Pulled By: gchanan

fbshipit-source-id: e3243a4e81ecb58373681df8bf6a00428352fb14

Ignore flake8 warning about whitespace before ':' (#15663)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15663

Ignore sometimes incorrect flake8 warning about whitespace before ':'

See https://github.com/ambv/black/issues/315

Reviewed By: soumith

Differential Revision: D13565818

fbshipit-source-id: 9d5ec2335899527ee71f4b505c00865a354e3bf0

Add count_include_pad arg for PoolOpGradient on CPU and fix ARM performance issue. (#15651)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15651

Add count_include_pad arg for PoolOpGradient on CPU and fix ARM performance issue.

Reviewed By: houseroad

Differential Revision: D13564257

fbshipit-source-id: 3a143f1122bc507ccb7827e9b46908d5c7203735

Unify the usage of Dequantize (#15685)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15685

The declaration of "Dequantize" is in "fbsource/fbcode/deeplearning/fbgemm2/QuantUtils.h", so it requires the "namespace fbgemm".

<T> is actually optional, since the type can de deduced from the first argument.

In some places we have "Dequantize<T>(...)", while in other places we have "Dequantize(...)". We'd better unify them. As a reference, all occurrences of "Quantize" are using "fbgemm::Quantize<T>(...)".

Reviewed By: jspark1105

Differential Revision: D13570847

fbshipit-source-id: 7fca9f7f9e4e0d9e5eb27ac44b8707adc3c80717

Fix vec256 inversion (#15659)

Summary:
soumith zou3519

I was browsing the code, and think `vec256_int.h` might need a minor revision, but not 100% sure.

1. It currently invert the result by `XOR` with 0. Should it `XOR` with 1 instead?
~2. AVX2 logical operations would set all bits in a byte/word/... to `1` if the condition holds. So functions, such as `_mm256_cmpeq_epi64 ` would return `0/-1` instead of `0/1`. Should it be masked with `1` to make sure it returns 0/1?~

~Would I be correct if I assume that the code revised below is not yet activated, but will be after we port legacy code to ATen?~
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15659

Differential Revision: D13565929

Pulled By: mrshenli

fbshipit-source-id: 8ae3daf256c3d915dd855a2215c95275e899ea8c

Add min/max on numbers to JIT

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15680

Differential Revision: D13568806

Pulled By: zdevito

fbshipit-source-id: ef0f33cc12a057184293bc31d28cc7b24f73eb94

initialize with ident value in global reduction (#15653)

Summary:
Fixes #15647. cc colesbury.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15653

Differential Revision: D13571132

Pulled By: soumith

fbshipit-source-id: 8f25943c974b3b931f4528e0e0a370bc095dab51

Updating submodules

Reviewed By: yns88

fbshipit-source-id: f7b540159cf1fe72825d09d55d56117d14ff90eb

Support for Jetson Xavier (#15660)

Summary:
The request changes are to support building Pytorch 1.0 on the Jetson Xavier with Openblas. Jetson Xavier with Jetpack 3.3 has generic lapack installed. To pick up the CUDA accelerated BLAS/Lapack, I had to build Openblas and build/link pytorch from source. Otherwise, I got a runtime error indicating lapack routines were not cuda enabled.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15660

Differential Revision: D13571324

Pulled By: soumith

fbshipit-source-id: 9b148d081d6e7fa7e1824dfdd93283c67f69e683

Fixing cuda100 smoke tests

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15673

Reviewed By: yf225

Differential Revision: D13568746

Pulled By: pjh5

fbshipit-source-id: e636de417d61b48074399da75bfb2576c9f62743

Remove PythonOp non-CPU path and PytorchOp (#15417)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15417

Right now the way we test whether Blob contains a CPU tensor is broken in ```PythonOpBase``` is broken, which means non-CPU path might never be taken.
Searching through the codebase, non-gpu path is used in PythonDLPack, and it is used in PytorchOp which is unused. So we'll remove non-gpu path in this diff.

Reviewed By: dzhulgakov

Differential Revision: D13495011

fbshipit-source-id: 9fe9537f05026d2a2cf7051efa81d184de722710

Updating submodules

Reviewed By: yns88

fbshipit-source-id: bb142e8f91046cc2b7ea32dac46ec0753b4bc218

fix select after chunk op (#15672)

Summary:
Fixes #15669.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15672

Differential Revision: D13567274

Pulled By: suo

fbshipit-source-id: a63e6cfc9dacedd4cb99dc51eee452038418001e

make flake8 failure blocking (#15675)

Summary:
Right now it just prints whatever flake8 errors and moves forward with the commit. This is too easy to miss.

It should block the commit so that the user can fix the issue
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15675

Differential Revision: D13567821

Pulled By: suo

fbshipit-source-id: 5f0de40ddd771bad8d6848417408cffbceb03183

redo sleef build fix (#15549)

Summary:
This was accidentally reverted by #14866
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15549

Differential Revision: D13549674

Pulled By: zdevito

fbshipit-source-id: e209aac53dccb082b91cfa2d292310eabeb459e3

format conv_test.py to prepare D13562099 (#15632)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15632

Just formatting and a few lints.

Reviewed By: yinghai

Differential Revision: D13562403

fbshipit-source-id: c56f8ee61f68cdaccc0828a764ff729454f68259

Fix torch.gesv args in doc

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15649

Differential Revision: D13564312

Pulled By: soumith

fbshipit-source-id: b3bba2ece600880077eb09b092ce17e331995bd6

clamp fixes (#15479)

Summary: fix to #15338 .

Differential Revision: D13564343

Pulled By: soumith

fbshipit-source-id: be64b572945533e10ae6f627d335b47f093720a3

Updating submodules

Reviewed By: cdelahousse

fbshipit-source-id: acb68439e62ea270af22364183a6ecba883fab66

Updating submodules

Reviewed By: cdelahousse

fbshipit-source-id: 5c5ad6a5cc9220ee1dd9565d64c7459f866ff74d

Fix typo in documentation

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15628

Differential Revision: D13562685

Pulled By: soumith

fbshipit-source-id: 1621fcff465b029142313f717035e935e9159513

Make btriunpack work for high dimensional batches and faster than before (#15286)

Summary:
Changelog:
- Optimize btriunpack by using `torch.where` instead of indexing, inplace operations instead of out place operations and avoiding costly permutations by computing the final permutation over a list.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15286

Differential Revision: D13562038

Pulled By: soumith

fbshipit-source-id: e2c94cfab5322bf1d24bf56d7b056619f553acc6

Add count_include_pad arg for average_pool_op on CPU (#15593)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15593

Add count_include_pad arg for average_pool_op on CPU

Reviewed By: houseroad

Differential Revision: D13558123

fbshipit-source-id: 188879ec3af313105ff66ac0b5a81ea44fca2855

Remove TH/THC link for cholesky (#15595)

Summary:
Changelog:
- Remove TH/THC binding
- Port single matrix case to ATen
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15595

Differential Revision: D13561657

Pulled By: soumith

fbshipit-source-id: 65f8c4b455cf19a0c7b6aeac2e3b985c7a7208f8

Concatenate directly into shared memory when constructing batches for numpy (#14534)

Summary:
Since #1323 tensors are shared with shared memory, but this feature is not active for numpy.
This PR fix this.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14534

Differential Revision: D13561649

Pulled By: soumith

fbshipit-source-id: b6bc9e99fb91e8b675c2ef131fba9fa11c1647c0

Add a patch for OSX with SDK<10.12 (#15615)

Summary:
Fixes https://github.com/pytorch/pytorch/issues/15614

Build passing on SDK 10.9
https://dev.azure.com/ramonaoptics/feedstock-builds/_build/results?buildId=13
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15615

Differential Revision: D13561737

Pulled By: soumith

fbshipit-source-id: 2ab0f78338d4949fa3f2735915fd96dce4bcd621

Fix typo: szie -> size

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15466

Differential Revision: D13536343

Pulled By: soumith

fbshipit-source-id: cb3df30bf346ef6bc0bc1b6430107b3e0e086f8d

Make the warning suppression safer (#15560)

Summary:
Address the problem introduced in https://github.com/pytorch/pytorch/pull/15499#issuecomment-450038494.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15560

Differential Revision: D13561346

Pulled By: soumith

fbshipit-source-id: 6abf622672bdcb77ae1a7188e8a3817fa97aecbc

add NCHW2NHWC and NHWC2NCHW in utils.py (#15588)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15588

Use NHWC2NCHW or NCHW2NHWC functions which is easier to understand compared to code using transpose and generalizable to non-2D convolutions.

Reviewed By: csummersea

Differential Revision: D13557674

fbshipit-source-id: c4fdb8850503ea58f6b17b188513ae2b29691ec0

Remove TH/THC link for gesv (#15510)

Summary:
This PR removes the TH/THC binding for gesv.

Changelog:
- Remove TH/THC binding
- Port single matrix case to ATen
- Enable test_gesv for CUDA as well
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15510

Differential Revision: D13559990

Pulled By: soumith

fbshipit-source-id: 9da2825e94d3103627e719709e6b1f8b521a07fb

keep extra_info of each op in ProfDagStats (#15244)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15244

This DIFF keeps track of the extra_info information attached to each operator. When getPerOpStas() is called, it attaches the extra_info to the result ProfDagStats protobuf.

Facebook
Net transform attaches a global_op_id which is defined as a tuple of (orig_net_name, original_op_index) to each operator,
The global_op_id is encoded as extra_info in each operator.

Reviewed By: aazzolini

Differential Revision: D13016289

fbshipit-source-id: 3e2719ec7ed0ebe47740b77581c565ff7e79b102

Error when torch.load-ing a JIT model (#15578)

Summary:
Throw a warning when calling `torch.load` on a zip file

Fixes #15570
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15578

Differential Revision: D13555954

Pulled By: driazati

fbshipit-source-id: a37ecdb3dd0c23eff809f86e2f8b74cd48ff7277

default_collate should collate bool list to byte tensors (#14669)

Summary:
Based on #15331 . Review only the last commit.

Fixes https://github.com/pytorch/pytorch/issues/14507.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14669

Reviewed By: ezyang

Differential Revision: D13528725

Pulled By: soumith

fbshipit-source-id: f12f1ac1c4ff2a3ddd6877c0c096a5da3a1ffa3c

append caffe2 prefix to dnnlowp cmd line options (#15582)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15582

Following convention of having caffe2_ prefix in command line options

Reviewed By: viswanathgs

Differential Revision: D13252055

fbshipit-source-id: 142a6395b832f211f34d0a87ec2d62c1e5fcdc69

adding nightly build smoke tests to circleci

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15441

Reviewed By: yf225

Differential Revision: D13552399

Pulled By: pjh5

fbshipit-source-id: 4a52ee2d08324b9ab6b8c266ad6a1cd3bdad1c71

add the int support (#15581)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15581

as title

Reviewed By: protonu

Differential Revision: D13556274

fbshipit-source-id: ba21f0970257d526e2fe7574eea4f89465b9c618