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

Implement rotated generate_proposals_op without opencv dependency (CPU version)

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18533

Reviewed By: ezyang

Differential Revision: D14648083

fbshipit-source-id: e53e8f537100862f8015c4efa4efe4d387cef551

Use SetOutputTensor instead of copying outputs manually (#17770)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17770

As title

Reviewed By: dzhulgakov

Differential Revision: D14370937

fbshipit-source-id: f415490c38556cf03bb13dce3643775331483448

Fix NCCL/Gloo process groups and DDP stream sync bug (#18465)

Summary:
DDP with NCCL backend uses a [worker stream](https://github.com/pytorch/pytorch/blob/d3eb941ed96774efb8d89a0b20c9e49807ea85a7/torch/csrc/distributed/c10d/ddp.cpp#L142) to flatten grand batch
tensors, and passes the flattened tensor to [another stream](https://github.com/pytorch/pytorch/blob/d3eb941ed96774efb8d89a0b20c9e49807ea85a7/torch/lib/c10d/ProcessGroupNCCL.cpp#L379) to
conduct ncclAllReduce. The flattened tensor has to record the
ncclAllReduce stream, otherwise multiple streams might access the
same memory space.

cc ppwwyyxx
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18465

Differential Revision: D14613449

Pulled By: mrshenli

fbshipit-source-id: b62773732552d12cc87b7adeb6897e9e11753ea9

Inference LSTM integration test (#18559)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18559

Adding integration test for inference LSTM

Reviewed By: houseroad

Differential Revision: D14656698

fbshipit-source-id: 80fb2a72be30fcb695f4471b72bf9d6e3965bf81

Add Slot type to abstract the raw pointers being used for slots. (#18226)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18226
ghimport-source-id: b9ec8651212875b30971cc6859d2ddec6559ae3a

If modules become first-class IValues, then the slots will no longer be raw pointers but (IValue, index) pairs. This commit inserts the Slot abstraction so that this change can be made in later patches.

Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18226 Add Slot type to abstract the raw pointers being used for slots.**

Differential Revision: D14542022

fbshipit-source-id: b81d7f4334c983d663e7551bda82df43680d7c5f

Revert D14635130: Improved onnx export for 3 onnx ops.

Differential Revision:
D14635130

Original commit changeset: d54a2b6e2950

fbshipit-source-id: f624e2befdde245cb88435a95508b2a8e6b12e61

Improved onnx export for 3 onnx ops. (#18512)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18512

Ceil and Floor have been supported since version 6 of ONNX: export them using the native onnx ops instead of an Aten op.
Similarly, support for the Where op has been added in version 9, so we don't need to wrap these op in an Aten op.

Reviewed By: houseroad

Differential Revision: D14635130

fbshipit-source-id: d54a2b6e295074a6214b5939b21051a6735c9958

Revert D14652372: [pytorch][PR] Add parsing to file check

Differential Revision:
D14652372

Original commit changeset: 7430b9d1dc2b

fbshipit-source-id: fa3d0f68515fe53447746469844d2db20c1292e0

C++17.h: forward -> c10::guts::forward (#18492)

Summary:
Use c10::guts::forward instead of forward
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18492

Reviewed By: smessmer

Differential Revision: D14625513

Pulled By: ilia-cher

fbshipit-source-id: 8bc4e20f102fe2a107a22f3e172882d60b95ab0e

Use __ldg for CUDA kernels in fuser (#18540)

Summary:
While benchmarking a kernel with broadcasted inputs, I noticed
that is was much slower than a hand-coded kernel for the smae task.

The kernel in question computed a * b + c for a of shape
32 x 32 x 10240 and b and c of shape 1 x 32 x 1.

This patch accellerates said kernel from 450us to 250us on my GTX1080Ti.

I didn't change half because there doesn't seem to be __ldg for
half.

An alternative could be to sprinkle const and restrict.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18540

Differential Revision: D14657840

Pulled By: soumith

fbshipit-source-id: 408847346ec12d1d1d9b119ac50bbc70f0d9ed33

Adds Cyclical Learning Rate and Momentum (#18001)

Summary:
This implements a cyclical learning rate (CLR) schedule with an optional inverse cyclical momentum. More info about CLR: https://github.com/bckenstler/CLR

This is finishing what #2016 started. Resolves #1909.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18001

Differential Revision: D14451845

Pulled By: sampepose

fbshipit-source-id: 8f682e0c3dee3a73bd2b14cc93fcf5f0e836b8c9

Completely synchronize behavior of Facebook flake8 and public flake8. (#18538)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18538
ghimport-source-id: 665b09f158d1c5dd94686d4212792504b55b7f73

Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18538 Completely synchronize behavior of Facebook flake8 and public flake8.**

Previously, developers at Facebook had the very funny experience
wherein /usr/local/bin/flake8 behaved differently than a freshly
installed flake8 from pip. In this commit, I add enough ignores to
.flake8 and install enough plugins to make the Facebook flake8
and public flake8 line up exactly. These means you don't have
to care which flake8 you use; they all will report accurate information
on your Python files.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14652336

fbshipit-source-id: ba7776eaa139cf2e3df2e65349da6fd7c99acca4

add slow tests annotation to some jit tests (#18545)

Summary:
Adds slow test annotation to the following very slow tests -

70.33s     test/test_jit.py::TestScript::test_script_module_script_resnet
32.33s     test/test_jit.py::TestBatched::test_beam_search
17.70s     test/test_jit.py::TestBatched::test_greedy_search
15.58s     test/test_jit.py::TestScript::test_script_module_trace_resnet18

The list of remaining slow tests is below. Let me know if you think any of the others should be added to slow tests as well. Slow tests will only run on master.

15.28s call     test/test_jit.py::TestJit::test_export_batchnorm
12.96s call     test/test_jit.py::TestEndToEndHybridFrontendModels::test_snli
11.65s call     test/test_jit.py::TestEndToEndHybridFrontendModels::test_neural_style
6.38s call     test/test_jit.py::TestJitGeneratedModule::test_nn_LocalResponseNorm_1d
5.96s call     test/test_jit.py::TestJitGeneratedModule::test_nn_LocalResponseNorm_2d_uneven_pad
5.91s call     test/test_jit.py::TestJitGeneratedModule::test_nn_LocalResponseNorm_3d_custom_params
4.76s call     test/test_jit.py::TestJit::test_alexnet
3.82s call     test/test_jit.py::TestScript::test_number_math
3.81s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_no_bias
3.76s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_groups_thnn
3.65s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_stride_pad1circular
3.49s call     test/test_jit.py::TestBatched::test_lstm
3.33s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_pad2circular
3.19s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv1d_stride1_pad2circular
3.11s call     test/test_jit.py::TestEndToEndHybridFrontendModels::test_dcgan_models
3.11s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_stride_padding
3.11s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_stride
3.08s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_no_bias
3.08s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv1d_stride1_pad1circular
3.07s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_groups
3.05s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_dilated
3.05s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_depthwise_with_multiplier
3.04s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_groups
3.03s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_dilated
3.02s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv2d_depthwise_dilated
3.02s call     test/test_jit.py::TestJitGeneratedModule::test_nn_Conv3d_dilated_strided
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18545

Differential Revision: D14656064

Pulled By: eellison

fbshipit-source-id: d17ee23c3b3679276cee983555d43e83ce099356

Add parsing to file check (#18304)

Summary:
This allows you to embed checks in IR, making the test more readable.

E.g.
```
graph_str = 'graph(%0 : Double(5, 5)):
          # CHECK: aten::relu
          %1 : Double(5, 5) = aten::relu(%0)
          return (%1)'
FileCheck().run(graph_str, parseIR(graph_str))
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18304

Differential Revision: D14652372

Pulled By: eellison

fbshipit-source-id: 7430b9d1dc2b7584704375aac02d7392ecec76a0

bug fix for node with writers in create autodiff subgraph (#18491)

Summary:
Previously we were moving nodes with writers into differentiable subgraphs, without necessarily preserving whether or not they were written to. This can lead to bugs with CSE, which needs that context.

I'm not completely sure if there's anything else we can do to be more aggresive here - inline these nodes and not run CSE and just run constant pooling, or possibly something else, but I think we should land this correctness condition first and then possibly think further.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18491

Differential Revision: D14648562

Pulled By: eellison

fbshipit-source-id: bc1e444774ccdb708e22f0e06a477a221a231f9e

add extra info for the auto gen sum ops

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/17934

Reviewed By: iroot900

Differential Revision: D14418689

fbshipit-source-id: 9e11e461001467f0000ea7c355d5b0f0d738fa85

Clarify error text of the pin_memory function

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18530

Reviewed By: ezyang

Differential Revision: D14647578

Pulled By: VitalyFedyunin

fbshipit-source-id: ddd70240d52d2e9a96e26f5a0dfea8d76fe25078

Move fast rnn benchmark to pytorch/pytorch

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18369

Differential Revision: D14652039

Pulled By: wanchaol

fbshipit-source-id: 1177b1f60d96672c3e2c9d527b56ee06ca7c0af1

Rename isTensor api -> isCompleteTensor (#18437)

Summary:
Is Tensor has been brought up as misleading a couple times, rename it isCompleteTensor for clarity.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18437

Differential Revision: D14605223

Pulled By: eellison

fbshipit-source-id: 189f67f12cbecd76516a04e67d8145c260c79036

Const trace error v2 (#18535)

Summary:
Trying to reland https://github.com/pytorch/pytorch/pull/18298
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18535

Differential Revision: D14652391

Pulled By: eellison

fbshipit-source-id: 699e30045dd5f14f0a2b98378272045a292e1e2a

enable more unit tests (#18537)

Summary:
Enable unit tests working with ROCm 2.3. In particular, these are unit tests where we skipped for double data types previously and some tests for multi-GPU setups.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18537

Differential Revision: D14651822

Pulled By: ezyang

fbshipit-source-id: 7dd575504ebe235a91489866c91000e9754b1235

Skip tests if C2/ONNX models cannot be read (#18494)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18494

Today we have some C2 end2end test run requiring reading model data from external filesystem (for example, Gluster and AWS). This could be a source for flaky test when the external filesystems are not reachable during the tests.

In this diff, we add try/catch logic around where we download models and open model files from external system. In case such attempts fails, we will catch the excption and let the unittest skip the current test instead of failure.

I also refactor the code a little bit by removing some duplicated logic on downloading and build the c2 model data. It has been duplicated in two classes and a few functions...

Reviewed By: yinghai

Differential Revision: D14442241

fbshipit-source-id: da8bf56c8d096efa34ca2070de5cd10a18aad70c

Add qtensors in caffe2 protobuf argument (#18486)

Summary:
We are about to merge onnxifi quantization support soon. Before that, I would like to merge this diff seperately to make sure it doesnt break anything.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18486

Reviewed By: bddppq, houseroad

Differential Revision: D14626419

Pulled By: yinghai

fbshipit-source-id: 504c1eae60be1e629203267b59defb8b69d82c0a

Generate sphinx docs with secure content. (#18508)

Summary:
There are a number of pages in the docs that serve insecure content. AFAICT this is the sole source of that.

I wasn't sure if docs get regenerated for old versions as part of the automation, or if those would need to be manually done.

cf. https://github.com/pytorch/pytorch.github.io/pull/177
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18508

Differential Revision: D14645665

Pulled By: zpao

fbshipit-source-id: 003563b06048485d4f539feb1675fc80bab47c1b

Fix loss functions doc (#18420)

Summary:
Correct docstring display error on web page caused by my previous PR
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18420

Differential Revision: D14642467

Pulled By: soumith

fbshipit-source-id: 16fdd3301a4c5bad27fbcd8686f7fbfcc1e908ee

Upgrade flake8-bugbear to master, fix the new lints. (#18507)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18507
ghimport-source-id: 1c3642befad2da78a7e5f39d6d58732b85c76267

Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18507 Upgrade flake8-bugbear to master, fix the new lints.**

It turns out Facebobok is internally using the unreleased master
flake8-bugbear, so upgrading it grabs a few more lints that Phabricator
was complaining about but we didn't get in open source.

A few of the getattr sites that I fixed look very suspicious (they're
written as if Python were a lazy language), but I didn't look more
closely into the matter.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14633682

fbshipit-source-id: fc3f97c87dca40bbda943a1d1061953490dbacf8

Add export annotations for functions in c10 (#18464)

Summary:
Fixes #18461.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18464

Differential Revision: D14620963

Pulled By: ezyang

fbshipit-source-id: c11f3967de2ac69c7140767c8fe73a85555e9f40

Back out "Revert D14613517: [pytorch][PR] Updating onnxtrt submodule to master branch" (#18514)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18514

Original commit changeset: d6267ddfc339

Reviewed By: bddppq

Differential Revision: D14634476

fbshipit-source-id: 2633b0b4c512d71001e5c20cd79c0c0d7856f942

update of fbcode/onnx to b29e78a4efb8e5d8995f576bbf19a846807829b6 (#18503)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18503

Previous import was c05f2ae412daf8fd64136ca354b97ccf73e0ea6c

Included changes:
- **[b29e78a4](https://github.com/onnx/onnx/commit/b29e78a4)**: update copyright for open governance (#1885) <Prasanth Pulavarthi>
- **[3b0ecd55](https://github.com/onnx/onnx/commit/3b0ecd55)**: open governance (#1881) <Prasanth Pulavarthi>
- **[bbe28349](https://github.com/onnx/onnx/commit/bbe28349)**: Revert "Adding Reverse op (#1804)" (#1882) <Lu Fang>
- **[5be3e223](https://github.com/onnx/onnx/commit/5be3e223)**: Adding Reverse op (#1804) <Peyman Manikashani>

Reviewed By: zrphercule

Differential Revision: D14632717

fbshipit-source-id: 2637a4090e7071a59caff3a910fa4f077906bf3c

Move weight offload inside backend construction functor (#18385)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18385

By moving the weight offload into the backend initialization function, we can instantiate the backend once by creating the OnnxifiOp once and then clean up the parameter workspace. And we need to keep hold of that instantiated net (OnnxifiOp) without cleaning it. Subsequent ctor of OnnxifiOp of the same model will hit the cached backend and they will not look into weight offloading, which is safe as the weight is already gone.

Reviewed By: ipiszy

Differential Revision: D14590379

fbshipit-source-id: f7f34016e09777ad3df0af487885cd14658e1044

fix #16448 (#18479)

Summary:
Fixes #16448

bddppq
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18479

Differential Revision: D14635360

Pulled By: ezyang

fbshipit-source-id: 4010319fbce050dd0bdf4da3cd1171b9737f3c4c

Add section about .code to docs

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18493

Differential Revision: D14634677

Pulled By: jamesr66a

fbshipit-source-id: 9ee065f6ce4218f725b93deb4c64b4ef55926145

how to use the `ccache` package on Ubuntu (#18495)

Summary:
Added full instructions for how to use the `ccache` package. Thanks.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18495

Differential Revision: D14635351

Pulled By: ezyang

fbshipit-source-id: 158e1052bae580e95f73644252fdbddcc0213128

Append c10 libs to TorchConfig.cmake (#18418)

Summary:
Fixes #18416.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18418

Differential Revision: D14635322

Pulled By: ezyang

fbshipit-source-id: 81cb658f73583e4cd0358173617f747ebf4f7f8a

Add some missing docs for tensor methods and attributes, new unittest to enforce tensors.rst no longer miss anything (#16057)

Summary:
This depend on https://github.com/pytorch/pytorch/pull/16039

This prevent people (reviewer, PR author) from forgetting adding things to `tensors.rst`.

When something new is added to `_tensor_doc.py` or `tensor.py` but intentionally not in `tensors.rst`, people should manually whitelist it in `test_docs_coverage.py`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16057

Differential Revision: D14619550

Pulled By: ezyang

fbshipit-source-id: e1c6dd6761142e2e48ec499e118df399e3949fcc

Revert D14613517: [pytorch][PR] Updating onnxtrt submodule to master branch

Differential Revision:
D14613517

Original commit changeset: dd20d718db55

fbshipit-source-id: d6267ddfc339d04f182e2de1750a601c8d6bf8c6

Fix direct comparison of OperatorDef proto structs (#18466)

Summary:
arguments order is okay to be different

ajyu
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18466

Differential Revision: D14627258

Pulled By: bddppq

fbshipit-source-id: 430e1fb1bea2c5639a547ae7c1652368788c86b9

Revert D14605905: [pytorch][PR] Add return_counts to torch.unique

Differential Revision:
D14605905

Original commit changeset: 555f5a12a8e2

fbshipit-source-id: c7874f5987893e956c022180a37763d88bba38db

Fix typo in Github links in elementwise_ops_schema.cc (#18018)

Summary:
s/elementwise_op_schema.cc/elementwise_ops_schema.cc
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18018

Differential Revision: D14612291

Pulled By: soumith

fbshipit-source-id: 09276283b9ff92c039ce530165c62cc8421fb443

Improve numerical precision of (s)logdet (#18449)

Summary:
Fixes https://github.com/pytorch/pytorch/issues/18448 and https://github.com/pytorch/pytorch/issues/18450
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18449

Differential Revision: D14611638

Pulled By: soumith

fbshipit-source-id: 4f1f27ab5316a92d2783e734169f599afed743cf

fix arange shape issue inconsistency across cpu and cuda (#18462)

Summary:
Fixes https://github.com/pytorch/pytorch/issues/18363
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18462

Differential Revision: D14620263

Pulled By: soumith

fbshipit-source-id: 223524cdda2f5d55c2ca8d4cdcf6f7a05a6c15eb

Updating onnxtrt submodule to master branch

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18441

Differential Revision: D14613517

Pulled By: bddppq

fbshipit-source-id: dd20d718db55942df9cce7acd1151d6902bc57ff

Minor fix for onnx ConstantOfShape export (#18199)

Summary:
Set value as tensor of 1 element instead of scalar, according to ONNX spec.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18199

Reviewed By: dzhulgakov

Differential Revision: D14542588

Pulled By: houseroad

fbshipit-source-id: 70dc978d870ebe6ef37c519ba4a20061c3f07372

Namedtuple return for solve, slogdet, sort, topk (#17093)

Summary:
More ops for https://github.com/pytorch/pytorch/issues/394. ~~Also need to rebase after landing #16186, because we need to update the whitelist of the new unit test added in #16186.~~

cc: ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17093

Differential Revision: D14620068

Pulled By: ezyang

fbshipit-source-id: deec5ffc9bf7624e0350c85392ee59789bad4237

Expose c10 operators to caffe2 by operator name (#18160)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18160

When exposing a c10 operator to the caffe2 frontend, don't use the operator schema but use the operator name instead.
This allows us to get rid of the existing mechanism for operator schema registration in a diff stacked on top.

Reviewed By: dzhulgakov

Differential Revision: D14513420

fbshipit-source-id: 6b08a9c6d9497eaf18b62361dd44bc07c7b4b76b

Test running a CUDA build on CPU machine. (#18242)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18242
ghimport-source-id: b949d312a48226a34f90304162e910acee7c95cd

Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18242 Test running a CUDA build on CPU machine.**
* #18362 Add ability to query if built with CUDA and MKL-DNN.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14584429

fbshipit-source-id: b54de5b33f0c795a7d9605d30576cdf9b74050fd

Properly use cudaGetLastError return code. (#18485)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18485

I don't know how (1) we landed the wrong version of the patch and (2) how
this passed the push blocking test

Reviewed By: pjh5

Differential Revision: D14621961

fbshipit-source-id: 0a3953d7adcdc79727a61c2acff65f436dcafe55

Move math::Axpy function to elementwise lib (#18316)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18316

Move math::Axpy function to elementwise lib

i-am-not-moving-c2-to-c10

Reviewed By: houseroad

Differential Revision: D14574697

fbshipit-source-id: 7cfbb2da295c8966c5328bd6b577cce2638eea62

Upgrade mkldnn to version 0.18.1 (#18463)

Summary:
Upgrade mkldnn to version 0.18.1
Fix the MKLDNN build issue if linking with MKL 2019.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18463

Differential Revision: D14620228

Pulled By: ezyang

fbshipit-source-id: 136074ad0e4631e1dde4ca1b0af4ee6a41e50913

Add Google tag (#17690)

Summary:
This PR adds a Global Site Tag to the site.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17690

Differential Revision: D14620816

Pulled By: zou3519

fbshipit-source-id: c02407881ce08340289123f5508f92381744e8e3

remove redundant --install_dir parameter in GEN_COMMAND (#18473)

Summary:
remove redundant --install_dir parameter in GEN_COMMAND, since "--install_dir parameter " already contained in ${GEN_COMMAND}.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18473

Differential Revision: D14620193

Pulled By: ezyang

fbshipit-source-id: ee9953b5d055f4b8beb3557f95f6539051b0028a

Resolving comments from Bool Tensor for CPU PR (#18165)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18165
ghimport-source-id: 55cb3fb63a25c2faab1725b4ec14c688bf45bd38

Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18166 Bool Tensor for CUDA
* **#18165 Resolved comments from Bool Tensor for CPU PR**
-------
------------
This is a follow up PR that resolves some additional feedback on one the of previous Bool Tensor PRs.

gchanan, here is a list of almost all the comments from the original PR with respective fixes and replies:

**[utils/python_scalars.h]** why is this converting from uint8_t and not bool? (comment?)
When i was adding this, i was testing by creating a tensor and then calling its .tolist(). it worked for bool and uint8_t equally good so i left uint8_t as thought it makes more sense as we are calling PyBool_FromLong. �Changing it to bool.

**[ATen/Dispatch.h]**better name?.
fixed.

**[test/test_torch.py]** what about other factories, such as full? (and more).
There is a test that goes through the factory methods - test_tensor_factories_empty. i added some bool cases above it and added a comment that once CUDA will be done, i will unite them and it will iterate not just between CUDA and CPU but also all types. ��Adding all bool cases now. Will unite in CUDA PR.

**[generic/THTensorMath.h]** any changes in this file actually needed?
Bad merge. Fixed.

**[TH/THTensor.h]** this generates code for random, clampedRandom, and cappedRandom -- do we have tests for all of these with bool?
Added

**[c10/core/ScalarType.h]** I'm not very confident about the lack of Bool here -- can you look at the call sites and see what makes sense to do here?
Added bool to the macro and created a similar one without for a single case which fails the build with errors:

_./torch/csrc/jit/symbolic_variable.h:79:20: error: ambiguous overload for ‘operator*’ (operand types are ‘const torch::jit::SymbolicVariable’ and ‘torch::jit::Value*’)
return (*this) * insertConstant(rhs);_

Differential Revision: D14605105

fbshipit-source-id: abf82d50e8f8c50b386545ac068268651b28496d

Unify cudaGetDeviceCount implementations. (#18445)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18445
ghimport-source-id: 30d018737bf6989bc68b7e3676f44e0ca6141fde

Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18242 Test running a CUDA build on CPU machine.
* **#18445 Unify cudaGetDeviceCount implementations.**

I went about doing this by searching for calls to cudaGetDeviceCount,
and then methodically replacing them with references to c10::cuda::device_count()
or at::cuda::device_count().

There is a point to doing this: the various implementations wildly differed
in their handling of what to do when cudaGetDeviceCount returns an error.
The final standardized behavior is that **all errors are swallowed** and
we return device count of zero. This indirectly fixes running CUDA builds
on CPU, which was broken in #17847.

I added 'noexcept' to the 'deviceCount' virtual method on DeviceGuardImpl.
This is a BC-breaking change for anyone inheriting from DeviceGuardImpl
but all you need to do is put 'noexcept' on your method and it is backwards
compatible with older libtorch.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14612189

fbshipit-source-id: 3c8d186e3dd623c0e27625212c7ce30f75d943cb

Use TensorIterator for unary operations

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18309

Differential Revision: D14591533

Pulled By: cpuhrsch

fbshipit-source-id: a3b0788a481bddf1803c9f2d3289263d7364f8d7

Introduce SobolEngine (#10505)

Summary:
`SobolEngine` is a quasi-random sampler used to sample points evenly between [0,1]. Here we use direction numbers to generate these samples. The maximum supported dimension for the sampler is 1111.

Documentation has been added, tests have been added based on Balandat 's references. The implementation is an optimized / tensor-ized implementation of Balandat 's implementation in Cython as provided in #9332.

This closes #9332 .

cc: soumith Balandat
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10505

Reviewed By: zou3519

Differential Revision: D9330179

Pulled By: ezyang

fbshipit-source-id: 01d5588e765b33b06febe99348f14d1e7fe8e55d

fix str of autogradzero

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18442

Differential Revision: D14602880

Pulled By: wanchaol

fbshipit-source-id: ebd00f9bb5f1f7e33964c10d8c9f165b7bb4985f

Optimize boolean expressions & unwraps (#18259)

Summary:
Simplify or eliminate boolean and/or expressions, optimize unwrapping a value that cannot be None, and optimize using `is` with a None and a non-None value

Since peephole optimize is now introducing constants, i added another constant propagation pass after running it.

Previously i had a PR that did this & optimized shape ops - i will add the shape optimizations in a separate PR.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18259

Differential Revision: D14602749

Pulled By: eellison

fbshipit-source-id: 1c3f5a67067d8dfdf55d7b78dcb616472ea8a267

Fix python resolution in caffe2 CI scripts

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18417

Differential Revision: D14612704

Pulled By: bddppq

fbshipit-source-id: 0942048a9c3990afc50ce73c1fa1005c4d4097aa

Support dim=None for argmax and argmin (#18264)

Summary:
Fixes: https://github.com/pytorch/pytorch/issues/18263
cc: houseroad
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18264

Reviewed By: ezyang

Differential Revision: D14559234

Pulled By: houseroad

fbshipit-source-id: c5b8623752d6c6af41c6d715fd9585a65294868d

Add return_counts to torch.unique (#18391)

Summary:
Fixes: https://github.com/pytorch/pytorch/issues/12598

This PR was originally authorized by ptrblck at https://github.com/pytorch/pytorch/pull/15495, but since there was no update for months after the request change, I clone that branch and resolve the code reviews here. Hope everything is good now. Especially, the implementation of count is changed from ptrblck's original algorithm to the one ngimel suggest, i.e. using `unique_by_key` and `adjacent_difference`.

The currently implementation of `_unique_dim` is VERY slow for computing inverse index and counts, see https://github.com/pytorch/pytorch/issues/18405. I will refactor `_unique_dim` in a later PR. For this PR, please allow me to keep the implementation as is.

cc: ptrblck ezyang ngimel colesbury
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18391

Reviewed By: soumith

Differential Revision: D14605905

Pulled By: VitalyFedyunin

fbshipit-source-id: 555f5a12a8e28c38b10dfccf1b6bb16c030bfdce

change dropout lowering in symbolic_script (#18375)

Summary:
Dropout is now eligible for fusion, and generated fused kernels are just as fast as dropout in ATen. Change its lowering in symbolic script so that it can actually be fused. Still special-cased for cuda, because without fusion this lowering is less efficient than current (bernoulli_ * input). Testing is covered by the test case that ailzhang added (test_dropout_cuda).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18375

Differential Revision: D14611938

Pulled By: soumith

fbshipit-source-id: 11b18f4784e6c9265e382a8f8deca7add8df3b37

Add torch.version.git_version (#18299)

Summary:
Fixes: https://github.com/pytorch/pytorch/issues/18293
cc: colesbury
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18299

Differential Revision: D14611972

Pulled By: soumith

fbshipit-source-id: cdb48ef37c8869713a9a43ea0da08e1bed9279a2

Change deprecated IntList to IntArrayRef

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18262

Differential Revision: D14612244

Pulled By: ezyang

fbshipit-source-id: 5d21c7b94d64104fececcb15c6d38d9bd2a1fc70

Enable printing to stderr for test_proper_exit for better debugging (#18458)

Summary:
related to https://github.com/pytorch/pytorch/issues/16608
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18458

Differential Revision: D14611718

Pulled By: soumith

fbshipit-source-id: 6dc903ff2d32b9c3b76470869d1f4e9a67f706df

Don't require pygraphviz for regenerate.sh (#17485)

Summary:
closes #17336

Do not overwrite config.yml if script throws an error
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17485

Differential Revision: D14604388

Pulled By: kostmo

fbshipit-source-id: 5024545e3a8711abdbc0800911c766929dbca196

Add quant-passes stubs. (#18151)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18151
ghimport-source-id: 7d12462971bdf3e5e26a3f150f1fcad05bba1a15

Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18152 Initial implementation of InsertObserverNodes pass.
* **#18151 Add quant-passes stubs.**

gh-metadata: pytorch pytorch 18149 gh/zolotukhinm@gmail.com/1/head

Differential Revision: D14584224

fbshipit-source-id: b3d0b5ff797160d5ad23f91f732e627b0129086c

caffe2 - support flaky operator tests for caffe2 build (#18155)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18155

- Make a python decorator caffe2_flaky for caffe2 operator unit tests.
- The environment variable CAFFE2_RUN_FLAKY_TESTS are now used to mark flaky test mode

During test run,
- If flaky tests mode are on, only flaky tests are run
- If flaky tests mode are off, only non-flaky tests are run

Mark ctc_beam_search_decoder_op_test as flaky

Reviewed By: ezyang, salexspb

Differential Revision: D14468816

fbshipit-source-id: dceb4a48daeb5437ad9cc714bef3343e9761f3a4

Remove unused th_scalar_type (#18390)

Summary:
th_scalar_type seems to be unused anywhere so can be removed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18390

Reviewed By: ezyang

Differential Revision: D14591374

Pulled By: izdeby

fbshipit-source-id: 2113aa81229cdfdfb8dc5c951ea6dea3725b8582

Porting CPU UpSample functions to ATen (#18020)

Summary:
This PR resolves partially #10482
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18020

Differential Revision: D14598029

Pulled By: ezyang

fbshipit-source-id: 513e7c6438ab6d5dc3f43241e7cb724744e9a287

Fix caffe2 build with BLAS=OpenBLAS (#18422)

Summary:
g++ complains about failing to find the declaration of cblas_sscal and cblas_dscal BLAS function
let's fix it :)

fedora 29, gcc 8.3.1, openblas 0.3.5
build with cmake -DBLAS=OpenBLAS ..
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18422

Differential Revision: D14598977

Pulled By: soumith

fbshipit-source-id: bde77bfb359d2ff38226401caeed78c114ef7468

Add addcmul, lerp to fuser, enable scalar->float specialization in symbolic script (#18081)

Summary:
This PR did two things:

1. Enable scalar->float specialization in symbolic script, so AD formula that contains scalar in the schema, should write `float` instead.
2. add addcmul, lerp to AD and fuser.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18081

Differential Revision: D14490493

Pulled By: wanchaol

fbshipit-source-id: b3b86d960d5f051b30733bc908b19786111cdaa4

Add ability to query if built with CUDA and MKL-DNN. (#18362)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18362
ghimport-source-id: 374b7ab97e2d6a894368007133201f510539296f

Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18242 Test running a CUDA build on CPU machine.
* **#18362 Add ability to query if built with CUDA and MKL-DNN.**

Fixes #18108.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14584430

fbshipit-source-id: 7605a1ac4e8f2a7c70d52e5a43ad7f03f0457473

Updating submodules

Reviewed By: yns88

fbshipit-source-id: b2c5eb7dfa9048e399461c00d1103e945a30a5bc

Implement reference counting for shared IPC CUDA tensors (#16854)

Summary:
This is to fix #16141 and similar issues.

The idea is to track a reference to every shared CUDA Storage and deallocate memory only after a consumer process deallocates received Storage.

ezyang Done with cleanup. Same (insignificantly better) performance as in file-per-share solution, but handles millions of shared tensors easily. Note [ ] documentation in progress.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16854

Differential Revision: D13994490

Pulled By: VitalyFedyunin

fbshipit-source-id: 565148ec3ac4fafb32d37fde0486b325bed6fbd1

Don't segfault on trying to get data_ptr of sparse tensor. (#18347)

Summary:
Also asserts in storage_initialized that there is a storage.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18347

Differential Revision: D14582028

Pulled By: gchanan

fbshipit-source-id: df3f5d181188f39e361839169fd054539c3b2839

Assert tensor isn't sparse in enforce_invariants. (#18338)

Summary:
There's no reason we can't check this, but I'm punting on implementing it for now. But it currently segfaults, so this is an improvements.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18338

Differential Revision: D14580308

Pulled By: gchanan

fbshipit-source-id: 44d4cafeab12e1beeb3453a2d4068d221c2e9c4f

Only look for Caffe2 package when shared (#18421)

Summary:
Previously it would look for the Config even if it was not written.

Fixed #18419
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18421

Differential Revision: D14597139

Pulled By: ezyang

fbshipit-source-id: c212cbf5dc91564c12d9d07e507c8285e11c6bdf

Add more options to the quantization model exporter (#18383)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18383

Add command line options for different quantization schemes.

Reviewed By: amylittleyang

Differential Revision: D14476862

fbshipit-source-id: 37fbf5b4c1c550121eae313f5a71d703a0a87f0f

Revert "Specialize optional tensor inputs to graphs in the JIT (#18360)" (#18411)

Summary:
This reverts commit 7cc7ed1322405ba3c627b9c5661a330f92c4183d.

I think it's better to sort out the issues raised in #18407 firs. I'm sorry for not stopping it earlier.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18411

Differential Revision: D14594937

Pulled By: soumith

fbshipit-source-id: 3c90b7fa7694e2f59e55607acecde4a47af801ea

Fix deprecated: type() -> scalar_type() (#18406)

Summary:
Sorry for not sending these fixes in a single PR. I found this compiler warning when I was working on something else, and I just go to GitHub and modify the file directly for convenience...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18406

Differential Revision: D14594180

Pulled By: soumith

fbshipit-source-id: 92f48513bc62fbe2c67c759d68830a973296e43b

Fix deprecated: type() -> scalar_type()

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18394

Differential Revision: D14593890

Pulled By: soumith

fbshipit-source-id: 92b9a8c22008341c0cc3b7a721bef1973c528daf

Added tensor size warning to F.mse_loss() (#18349)

Summary:
To address the issue of broadcasting giving the wrong result in `nn.MSELoss()` as mentioned here https://github.com/pytorch/pytorch/issues/16045 . In particular, the issue often arises when computing the loss between tensors with shapes (n, 1) and (n,)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18349

Differential Revision: D14594176

Pulled By: soumith

fbshipit-source-id: f23ae68a4bf42f3554ad7678a314ba2c7532a6db

Fix For Requires Grad Infinite Loop (#18361)

Summary:
Previously, we would continue to run requires grad on a loop body when the outputs and inputs disagreed. This adds a check so that we don't continue running if the results haven't changed since the last run.

Fix for https://github.com/pytorch/pytorch/issues/18320
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18361

Differential Revision: D14584332

Pulled By: eellison

fbshipit-source-id: 696b225f80a2036318540946428b525985a9e735

update magma instructions (#18410)

Summary:
fixes https://github.com/pytorch/pytorch/issues/18389

cc: stas00
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18410

Differential Revision: D14594198

Pulled By: soumith

fbshipit-source-id: fb46ef77a36c90ad95e47f7066f5d32aa1f1370f

Removed some dead code (#18201)

Summary:
Removed some dead code.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18201

Differential Revision: D14555251

Pulled By: izdeby

fbshipit-source-id: f49640133ef4ae1b0306f7cec6655f23869cc6e7

Specialize optional tensor inputs to graphs in the JIT (#18360)

Summary:
This specializes optional tensor inputs to either a DimensionedTensorType or, when None is passed,
UndefinedTensor (aka AutogradZeroTensorType).
This works because we already have different specs and thus separate plans for the two cases.
It enhances the shape analysis - because now unwrapped optional tensors will have DimensionedTensorType with appropriate shape and required grad etc.
Also, when combined with "if-pruning" (which I understand #18259 works towards), we actually get much nicer concrete graphs, too.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18360

Differential Revision: D14590577

Pulled By: soumith

fbshipit-source-id: cac204a506d1d38b15703cbcc67a6b75fd4979f4

Move pyobj_ to TensorImpl (#18225)

Summary:
Currently, `THPVariable_Wrap(…)` and `THPVariable_NewWithVar(…)` depend on the existence of `pyobj_` in the autograd metadata of a Variable to convert the Variable to a Python tensor. However, after the Variable/Tensor merge, there will be Variables that don't contain autograd metadata, and to allow the conversion from non-autograd-meta Variable to a Python tensor we need to store the `pyobj_` outside of autograd metadata and in a place where it will always be available.

This PR makes it possible by moving `pyobj_` into TensorImpl, so that `THPVariable_Wrap(…)` and `THPVariable_NewWithVar(…)` can always access a Variable's `pyobj_` and convert the Variable to a Python tensor.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18225

Differential Revision: D14562616

Pulled By: yf225

fbshipit-source-id: 18d4aaace70eee6120abaf9276036d1f8f51b18d

Fix deprecated scalar type in ATen/native/Distributions.cpp

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18265

Differential Revision: D14577543

Pulled By: ezyang

fbshipit-source-id: 36674530b32366c51835e4073d7ba23d455d2fda

Revert D14446895: [C2] Implement rotated generate_proposals_op without opencv dependency (~2x faster)

Differential Revision:
D14446895

Original commit changeset: 847f2443e645

fbshipit-source-id: fc6ab5ee59e027f125f5ab0f7ee51ad7db37d4a4

Revert D14584266: [pytorch][PR] Better error message for tensor with grad as constant in tracing

Differential Revision:
D14584266

Original commit changeset: 4e7850dadc78

fbshipit-source-id: 3bb3b5006e469edff984c16e0ff8d5dac2862d88

Better error when module attr is used (#18164)

Summary:
Adds a suggestion to add to __constants__ when a torch.nn.Module attr is accessed
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18164

Differential Revision: D14580060

Pulled By: eellison

fbshipit-source-id: 0c5adc21d7341a5691d4b45930947cb1ba84c8e8

Fix incorrect sparse add behavior when the sparse tensor has non-contiguous values (#18179)

Summary:
Currently, this code gives incorrect result:
```python
import torch
indices=torch.tensor([[7, 1, 3]])
values=torch.tensor([[1., 1., 1.],
               [1., 1., 1.],
               [1., 1., 1.]])
x = torch.sparse_coo_tensor(indices, values, size=(10, 3))
values=torch.tensor(1.).expand(3, 3)
y = torch.sparse_coo_tensor(indices, values, size=(10, 3))
z = x + y

tensor(indices=tensor([[7, 1, 3]]),
       values=tensor([[2., 1., 1.],
                      [1., 1., 1.],
                      [1., 1., 1.]]),
       size=(10, 3), nnz=3, layout=torch.sparse_coo)
```

This PR fixes the bug by adding special handling for sparse tensors with non-contiguous values in the addition function (specifically, by cat'ing the indices and values together).

This PR closes https://github.com/pytorch/pytorch/issues/17950 and https://github.com/pytorch/pytorch/issues/17919.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18179

Reviewed By: ezyang

Differential Revision: D14569591

Pulled By: yf225

fbshipit-source-id: f5a14c4a31337fc95eab64596212066b4fb18b1a

Implement rotated generate_proposals_op without opencv dependency (1.8x faster) (#18010)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18010

[C2] Implement rotated generate_proposals_op without opencv dependency.

Reviewed By: newstzpz

Differential Revision: D14446895

fbshipit-source-id: 847f2443e645f8cae1327dfbaa111c48875ca9be

Remove empty file (actual file_check.cpp resides in torch/csrc/jit/testing) (#18303)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18303
ghimport-source-id: 66f4402075b123e36c6ffdf806b7c93187a1a58a

Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18307 Convert test_recursive_cse to use Filecheck inline annotations.
* #18306 [Filecheck] Add a feature to parse check annotations from string.
* #18305 Add python bindings for parseIR.
* **#18303 Remove empty file (actual file_check.cpp resides in torch/csrc/jit/testing)**

Differential Revision: D14586003

fbshipit-source-id: a13e57bd4302e4d3f06198068d525de25e2aa8b3

Turn script_type_parser into a class (#18211)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18211
ghimport-source-id: 73b81e9ec631937b14db1da10991831788a6894b

Stack from [ghstack](https://github.com/ezyang/ghstack):
* #18296 [jit] Add namespacing for ScriptClasses
* #18284 [jit] make test module hook use save/load
* **#18211 [jit] Turn script_type_parser into a class**
* #18148 [jit] python interop for script classes

If we are namespacing classes, the type parser will need to carry around
some state about which namespaces to look in. This PR just wraps it in a
class in preparation.

Also, subscriptToType can no longer be static, since parseTypeFromExpr
may give different results depending on the namespaces available, so
it's been made a regular function instead of a static map lookup.

Reviewed By: eellison

Differential Revision: D14581128

fbshipit-source-id: 711315472ccde1920abf9fdb5a871ac27fb86787

python interop for script classes (#18148)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18148
ghimport-source-id: 40a9d745dc9aeba53d098743323fcbd50ca65137

Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18148 py interop**

Support for converting classes between the Python–TorchScript boundary. Like other TorchScript values, ScriptClasses are native Python values when used in Python and IValues when used in TorchScript.

Notably, there is a copy across this boundary, which will be surprising to users who will expect standard Python reference semantics. I have some ideas for fixing that, but it's a more involved process.

Reviewed By: jamesr66a

Differential Revision: D14526259

fbshipit-source-id: 5916e3032488a42dc7da756c1826d7c040a21ebd

Better error message for tensor with grad as constant in tracing (#18298)

Summary:
Fix for https://github.com/pytorch/pytorch/issues/17583

There's an unrelated issue right now causing a segfault when printing tensor so that might have to fixed first for this to land
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18298

Differential Revision: D14584266

Pulled By: eellison

fbshipit-source-id: 4e7850dadc78ef1e98ad40b9d8adc0fef42acf48

Support for basic list comprehensions (#17267)

Summary:
Supports the following syntax:
```
        torch.jit.script
        def comp(l):
            # type: (List[float]) -> List[float]

            n = [x * 3 for x in l]
            return n
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17267

Differential Revision: D14581119

Pulled By: Krovatkin

fbshipit-source-id: 6fd091a8a9ab607386ac58fda6ad88bf8aea380e

Make it possible to trigger XLA/slow tests via commit message. (#18345)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18345
ghimport-source-id: 9649d76bb194866859d62e6ba2a3a265c96ebba5

Stack from [ghstack](https://github.com/ezyang/ghstack):
* **#18345 Make it possible to trigger XLA/slow tests via commit message.**

Four variants are supported: `[xla ci] [ci xla] [xla test] [test xla]`; substitute
xla with slow for slow tests.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Differential Revision: D14584557

fbshipit-source-id: fcbfdfb28246823135bb3d3910baae073d16e81d

Avoid refcount when looking up dispatch key

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18294

Reviewed By: ezyang

Differential Revision: D14512979

fbshipit-source-id: 45e548974f06184c375c2bb8339e3049a4ebd880