platform/upstream/pytorch.git
5 years agoSupport for Jetson Xavier (#15660)
rtarquini [Thu, 3 Jan 2019 02:48:31 +0000 (18:48 -0800)]
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

5 years agoFixing cuda100 smoke tests
Jesse Hellemn [Thu, 3 Jan 2019 01:10:35 +0000 (17:10 -0800)]
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

5 years agoRemove PythonOp non-CPU path and PytorchOp (#15417)
Jerry Zhang [Thu, 3 Jan 2019 00:32:02 +0000 (16:32 -0800)]
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

5 years agoUpdating submodules
svcscm [Wed, 2 Jan 2019 22:55:43 +0000 (14:55 -0800)]
Updating submodules

Reviewed By: yns88

fbshipit-source-id: bb142e8f91046cc2b7ea32dac46ec0753b4bc218

5 years agofix select after chunk op (#15672)
Michael Suo [Wed, 2 Jan 2019 22:32:00 +0000 (14:32 -0800)]
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

5 years agomake flake8 failure blocking (#15675)
Michael Suo [Wed, 2 Jan 2019 20:50:13 +0000 (12:50 -0800)]
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

5 years agoredo sleef build fix (#15549)
Zachary DeVito [Wed, 2 Jan 2019 20:45:38 +0000 (12:45 -0800)]
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

5 years agoformat conv_test.py to prepare D13562099 (#15632)
Jongsoo Park [Wed, 2 Jan 2019 19:25:41 +0000 (11:25 -0800)]
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

5 years agoFix torch.gesv args in doc
kiendang [Wed, 2 Jan 2019 08:18:07 +0000 (00:18 -0800)]
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

5 years agoclamp fixes (#15479)
surgan12 [Wed, 2 Jan 2019 07:09:45 +0000 (23:09 -0800)]
clamp fixes (#15479)

Summary: fix to #15338 .

Differential Revision: D13564343

Pulled By: soumith

fbshipit-source-id: be64b572945533e10ae6f627d335b47f093720a3

5 years agoUpdating submodules
svcscm [Wed, 2 Jan 2019 03:41:31 +0000 (19:41 -0800)]
Updating submodules

Reviewed By: cdelahousse

fbshipit-source-id: acb68439e62ea270af22364183a6ecba883fab66

5 years agoUpdating submodules
svcscm [Wed, 2 Jan 2019 01:20:19 +0000 (17:20 -0800)]
Updating submodules

Reviewed By: cdelahousse

fbshipit-source-id: 5c5ad6a5cc9220ee1dd9565d64c7459f866ff74d

5 years agoFix typo in documentation
Alexander Rodin [Mon, 31 Dec 2018 02:05:29 +0000 (18:05 -0800)]
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

5 years agoMake btriunpack work for high dimensional batches and faster than before (#15286)
vishwakftw [Sun, 30 Dec 2018 20:39:10 +0000 (12:39 -0800)]
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

5 years agoAdd count_include_pad arg for average_pool_op on CPU (#15593)
Xiaomeng Yang [Sun, 30 Dec 2018 12:13:54 +0000 (04:13 -0800)]
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

5 years agoRemove TH/THC link for cholesky (#15595)
vishwakftw [Sun, 30 Dec 2018 01:50:32 +0000 (17:50 -0800)]
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

5 years agoConcatenate directly into shared memory when constructing batches for numpy (#14534)
Christoph [Sun, 30 Dec 2018 01:48:36 +0000 (17:48 -0800)]
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

5 years agoAdd a patch for OSX with SDK<10.12 (#15615)
Mark Harfouche [Sun, 30 Dec 2018 00:09:12 +0000 (16:09 -0800)]
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

5 years agoFix typo: szie -> size
Gao, Xiang [Sat, 29 Dec 2018 06:38:24 +0000 (22:38 -0800)]
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

5 years agoMake the warning suppression safer (#15560)
peter [Sat, 29 Dec 2018 06:10:08 +0000 (22:10 -0800)]
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

5 years agoadd NCHW2NHWC and NHWC2NCHW in utils.py (#15588)
Jongsoo Park [Sat, 29 Dec 2018 01:32:11 +0000 (17:32 -0800)]
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

5 years agoRemove TH/THC link for gesv (#15510)
Vishwak Srinivasan [Sat, 29 Dec 2018 00:51:45 +0000 (16:51 -0800)]
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

5 years agokeep extra_info of each op in ProfDagStats (#15244)
Dong Li [Fri, 28 Dec 2018 23:00:41 +0000 (15:00 -0800)]
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

5 years agoError when torch.load-ing a JIT model (#15578)
David Riazati [Fri, 28 Dec 2018 21:52:01 +0000 (13:52 -0800)]
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

5 years agodefault_collate should collate bool list to byte tensors (#14669)
SsnL [Fri, 28 Dec 2018 19:51:26 +0000 (11:51 -0800)]
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

5 years agoappend caffe2 prefix to dnnlowp cmd line options (#15582)
Jongsoo Park [Fri, 28 Dec 2018 19:49:22 +0000 (11:49 -0800)]
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

5 years agoadding nightly build smoke tests to circleci
Jesse Hellemn [Fri, 28 Dec 2018 18:44:47 +0000 (10:44 -0800)]
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

5 years agoadd the int support (#15581)
Lingyi Liu [Fri, 28 Dec 2018 01:13:50 +0000 (17:13 -0800)]
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

5 years agoMove VariableImpl functions to AutogradMeta and Variable (#15487)
Will Feng [Fri, 28 Dec 2018 01:12:27 +0000 (17:12 -0800)]
Move VariableImpl functions to AutogradMeta and Variable (#15487)

Summary:
In this PR, we are moving all functions away from `Variable::Impl`, in order to get rid of `Variable::Impl` (and the `data_` Tensor in it) in the next PR. Some of the functions (such as `set_requires_grad` / `requires_grad` / `grad`) will be living in `AutogradMeta` class, while others (such as `backward()` / `rebase_history()` / `grad_accumulator()` / `grad_fn()`) will be living in `Variable` class.

This is the 2nd PR mentioned in https://github.com/pytorch/pytorch/issues/13638.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15487

Differential Revision: D13553173

Pulled By: yf225

fbshipit-source-id: 691f9432d0cd0640af380c757f3e3a2f64f8851c

5 years agotest basic tensor interop
Roy Li [Fri, 28 Dec 2018 01:01:19 +0000 (17:01 -0800)]
test basic tensor interop

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/12249

Differential Revision: D13469356

Pulled By: li-roy

fbshipit-source-id: b49748462aa44ac34b8ce79783f2c895a537a232

5 years agoAllow int/float cast to bool (#13391)
David Riazati [Thu, 27 Dec 2018 23:58:32 +0000 (15:58 -0800)]
Allow int/float cast to bool (#13391)

Summary:
This PR adds explicit `bool()` casts to match Python semantics

`bool(1) = True`
`bool(0) = False`
`bool(0.0) = False`
`bool(0.1) = True`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13391

Differential Revision: D12871213

Pulled By: driazati

fbshipit-source-id: 773a48b2647973138efe854abe725d647f1d727d

5 years agoremove print ops before exporting onnx graph (#15550)
Elias Ellison [Thu, 27 Dec 2018 23:35:24 +0000 (15:35 -0800)]
remove print ops before exporting onnx graph (#15550)

Summary:
Removing print ops before exporting onnx graph, fixes https://github.com/pytorch/pytorch/issues/15505
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15550

Differential Revision: D13551195

Pulled By: eellison

fbshipit-source-id: 1ea1e34cb5b8433eacc2b86fb10b241198af96be

5 years agoAdded deviceCount() virtual method to DeviceGuardImplInterface (#15574)
Igor Fedan [Thu, 27 Dec 2018 23:24:22 +0000 (15:24 -0800)]
Added deviceCount() virtual method to DeviceGuardImplInterface (#15574)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15574

Added deviceCount() virtual method to DeviceGuardImplInterface, also added correspondent implementation for CPUGuardImpl, CUDAGuardImpl, FakeGuardImpl, VirtualGuardImpl, HIPGuardImplMasqueradingAsCUDA

Reviewed By: soumith

Differential Revision: D13554609

fbshipit-source-id: 913bf2aad44a0a356efe54505ee4abaf6c4622db

5 years agoPort torch.range to aten and parallelize on CPU.
Gregory Chanan [Thu, 27 Dec 2018 23:20:42 +0000 (15:20 -0800)]
Port torch.range to aten and parallelize on CPU.

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15484

Differential Revision: D13538955

Pulled By: gchanan

fbshipit-source-id: ee3889ad116988d963e603621310b3bbdce0aec9

5 years agoExport group norm as ATen and add test (#15569)
Lu Fang [Thu, 27 Dec 2018 22:42:01 +0000 (14:42 -0800)]
Export group norm as ATen and add test (#15569)

Summary:
Short term solution, export group norm as an ATen op to unblock users.
Long term will add GroupNorm to onnx.

Add an end to end test for this one.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15569

Differential Revision: D13554293

Pulled By: houseroad

fbshipit-source-id: b4974c9ea2a1b81338ca1e5c6747efe2715d7932

5 years agoUpdate cuda.get/set_rng_state doc (#14324)
SsnL [Thu, 27 Dec 2018 22:06:23 +0000 (14:06 -0800)]
Update cuda.get/set_rng_state doc (#14324)

Summary:
Now that `cuda.get/set_rng_state` accept `device` objects, the default value should be an device object, and doc should mention so.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14324

Reviewed By: ezyang

Differential Revision: D13528707

Pulled By: soumith

fbshipit-source-id: 32fdac467dfea6d5b96b7e2a42dc8cfd42ba11ee

5 years agoUpdate QNNPACK (#15561)
Marat Dukhan [Thu, 27 Dec 2018 19:55:02 +0000 (11:55 -0800)]
Update QNNPACK (#15561)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15561

- Update QNNPACK submodule to master (API-incompatible)
- Do matching changes in Caffe2 Int8 operators

Reviewed By: dreiss

Differential Revision: D13551322

fbshipit-source-id: 066f9087061167f7d7cfbc1c8f8628dfa93d056e

5 years agoRevert D13552080: [pytorch][PR] add clang-format check to CI
Michael Suo [Thu, 27 Dec 2018 18:53:58 +0000 (10:53 -0800)]
Revert D13552080: [pytorch][PR] add clang-format check to CI

Differential Revision:
D13552080

Original commit changeset: 462a73894c16

fbshipit-source-id: ebfc5aa3343cebabbc24ff39e4e9841a372443e2

5 years agoFix wrong class name in jit _make_fail (#15559)
daquexian [Thu, 27 Dec 2018 09:59:56 +0000 (01:59 -0800)]
Fix wrong class name in jit _make_fail (#15559)

Summary:
It should be ScriptModule rather than TracedModule :)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15559

Differential Revision: D13552058

Pulled By: soumith

fbshipit-source-id: 0aa17639c225818b00d59daec4bc2336f039f658

5 years agoadd clang-format check to CI (#15543)
Michael Suo [Thu, 27 Dec 2018 06:17:59 +0000 (22:17 -0800)]
add clang-format check to CI (#15543)

Summary:
Simple check that runs against your PR's changes and complains if running clang-format would have created a change. Does nothing when run against master, so it's "safe" to accept changes that fail this check and it won't break the build.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15543

Reviewed By: soumith

Differential Revision: D13552080

Pulled By: suo

fbshipit-source-id: 462a73894c16e7108806af7fa88440c377d4d0d2

5 years agoFix github branch prefix v (#15552)
Ailing Zhang [Thu, 27 Dec 2018 03:43:10 +0000 (19:43 -0800)]
Fix github branch prefix v (#15552)

Summary:
Fixes #15519 .
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15552

Differential Revision: D13550780

Pulled By: ailzhang

fbshipit-source-id: b117e5ced42de207b91045bffcee8907dd73201e

5 years agoRotated boxes support for GPU GenerateProposals op (#15470)
Viswanath Sivakumar [Thu, 27 Dec 2018 02:01:20 +0000 (18:01 -0800)]
Rotated boxes support for GPU GenerateProposals op (#15470)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15470

On top of D13509114 and D13017791. Pretty straight-forward.

Reviewed By: newstzpz

Differential Revision: D13536671

fbshipit-source-id: ff65981b70c63773ccc9aef3ff28e3c9508f6716

5 years agoCUDA kernel for rotated NMS support, over 200x speedup than CPU (#15365)
Viswanath Sivakumar [Thu, 27 Dec 2018 02:01:20 +0000 (18:01 -0800)]
CUDA kernel for rotated NMS support, over 200x speedup than CPU (#15365)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15365

On top of D13017791, adding rotated NMS support with the same kernel building
blocks. Results in 218x speedup on avg.

Reviewed By: SuperIRabbit

Differential Revision: D13509114

fbshipit-source-id: c1d33c8dc4bc50b5906b4f01bb0caf1115e2a357

5 years agoMove autograd metadata from VariableImpl to TensorImpl (#13827)
Will Feng [Thu, 27 Dec 2018 00:31:47 +0000 (16:31 -0800)]
Move autograd metadata from VariableImpl to TensorImpl (#13827)

Summary:
Changes originally in this PR:
1. Move Variable::Impl data members into TensorImpl as `AutogradMeta` struct
2. Change Variable::Impl functions to use data members in `AutogradMeta` struct
3. Add `shallow_copy_and_detach()` function to each subclass of TensorImpl
4. Do shallow copy when the user calls `make_variable(tensor)` / `make_variable_view(tensor)` / `variable.set_data(tensor)` / `variable.detach()`

Changes moved from https://github.com/pytorch/pytorch/pull/13645:
1. Add a flag to Variable to disallow size/stride/storage_ptr changes from in-place operations such as `resize_` / `resize_as_` / `set_` / `transpose_`, and set this flag to true when people call `tensor.data` in Python.
2. Write text in the docs to actively discourage changing the shape or storage of `tensor_detached` and expecting `tensor` to also be updated.

This is the 1st+2nd PR mentioned in https://github.com/pytorch/pytorch/issues/13638.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13827

Differential Revision: D13507173

Pulled By: yf225

fbshipit-source-id: b177b08438d534a8197e34e1ad4a837e2db0ed6a

5 years agoversion bump to 1.1 (#15554)
Soumith Chintala [Wed, 26 Dec 2018 23:41:46 +0000 (15:41 -0800)]
version bump to 1.1 (#15554)

Summary:
version bump to 1.1
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15554

Differential Revision: D13550818

Pulled By: soumith

fbshipit-source-id: 8a28582c98b42c081e103581551a01fd96c9f42d

5 years agoIn README.md CMAKE_PREFIX_PATH should be CONDA_PREFIX when using an conda virtual...
derek [Wed, 26 Dec 2018 20:54:17 +0000 (12:54 -0800)]
In README.md CMAKE_PREFIX_PATH should be CONDA_PREFIX when using an conda virtual environment (#15548)

Summary:
In current README.md, `CMAKE_PREFIX_PATH` is set to conda root even when you have activated an virtual environment. When an conda virtualenv is activated, packages are installed in `CONDA_PREFIX`, not conda root. I think `CMAKE_PREFIX_PATH` should also be set to `CONDA_PREFIX` in this case. I think some build issues can be solved with the new instruction. Maybe something like #14954.

soumith,
When I made PR #15335 I was confused and made a wrong point. I think this PR could be the real solution.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15548

Differential Revision: D13549681

Pulled By: soumith

fbshipit-source-id: 42d855b6e49ee58d735d2f4715d3e5752a748693

5 years agoadd from_pretrained method to EmbeddingBag (#15273)
David Pollack [Wed, 26 Dec 2018 16:31:00 +0000 (08:31 -0800)]
add from_pretrained method to EmbeddingBag (#15273)

Summary:
The `EmbeddingBag` module does not include a `from_pretrained` method like the `Embedding` module.  I added it for consistency between the two modules.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15273

Differential Revision: D13547842

Pulled By: soumith

fbshipit-source-id: 8ffde51ff0c1e8fc8310263b6f375da88089ff7d

5 years agoMake argument size checking consistent across CPU and CUDA for torch.gesv (#15430)
vishwakftw [Wed, 26 Dec 2018 16:29:57 +0000 (08:29 -0800)]
Make argument size checking consistent across CPU and CUDA for torch.gesv (#15430)

Summary:
There is an inconsistency in the size of arguments for gesv, which is fixed in this PR.

Changelog:
- Replicate check in CPU as done for CUDA
- Fix argument ordering (minor) in CUDA checking

Fixes #15328

Differential Revision: D13531167

Pulled By: soumith

fbshipit-source-id: c4b4e4fc12880208d08e88d1e47e730ac98c2ad3

5 years agoclang format world (#15524)
Michael Suo [Wed, 26 Dec 2018 14:52:25 +0000 (06:52 -0800)]
clang format world (#15524)

Summary:
The PR clang-formats everything in `torch/csrc/jit/` and adds it to the pre-commit hook.

Here is a list of non-mechanical changes:
- I went over each file and fixed up whenever I could tell that clang-format was clobbering comment formatting.
- Made the macros in register_prim_ops a little more clang-format friendly by omitting trailing commas
- Refactored autodiff.cpp to use a helper class with explicit state rather than a bunch of capturing lambdas
- Small improvements to the precommit hook clang-format
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15524

Differential Revision: D13547989

Pulled By: suo

fbshipit-source-id: 3ff1541bb06433ccfe6de6e33f29227a2b5bb493

5 years agoAdded correct isinf handling for Integral tensors (#15489)
Frank Zhang [Wed, 26 Dec 2018 14:32:44 +0000 (06:32 -0800)]
Added correct isinf handling for Integral tensors (#15489)

Summary:
Currently torch.isinf on integral tensor will raise RuntimeError: value cannot be converted to type int16_t without overflow: inf.
This pr will suppress the error and return false(0) for all integral tensors. The behavior will also be consistent with np.isinf
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15489

Reviewed By: zou3519

Differential Revision: D13540786

Pulled By: flashhack

fbshipit-source-id: e730dea849da6a59f3752d347bcfbadfd12c6483

5 years agoTrivial comment update in autograd/function.h (#15529)
Derek Kim [Wed, 26 Dec 2018 10:11:17 +0000 (02:11 -0800)]
Trivial comment update in autograd/function.h (#15529)

Summary:
I removed the explanation on `num_inputs` parameter. This parameter was removed in #8168

colesbury
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15529

Differential Revision: D13547854

Pulled By: soumith

fbshipit-source-id: 8a9ac58f2c93a2533b82ec63089477166ed0bcb9

5 years agoFix failed type cast in Windows Debug Build (#15333)
peter [Wed, 26 Dec 2018 08:46:13 +0000 (00:46 -0800)]
Fix failed type cast in Windows Debug Build (#15333)

Summary:
Fixes #15330
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15333

Differential Revision: D13531317

Pulled By: soumith

fbshipit-source-id: b956f27bd7fa33cbdf405338fcbcbc7df2fd629f

5 years agoUpgrade MKL-DNN to version 0.17 and static build MKL-DNN (#15504)
Gu, Jinghui [Wed, 26 Dec 2018 06:54:16 +0000 (22:54 -0800)]
Upgrade MKL-DNN to version 0.17 and static build MKL-DNN (#15504)

Summary:
Upgrade MKl-DNN to 0.17 and static build MKL-DNN to fix the potentail build error due to old mkldnn version in host system.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15504

Differential Revision: D13547885

Pulled By: soumith

fbshipit-source-id: 46f790a3d9289c1e153e51c62be17c5206ea8f9a

5 years agoremove legacy from docs (#15112)
Soumith Chintala [Wed, 26 Dec 2018 05:55:26 +0000 (21:55 -0800)]
remove legacy from docs (#15112)

Summary:
Fixes https://github.com/pytorch/pytorch/issues/15062
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15112

Differential Revision: D13547845

Pulled By: soumith

fbshipit-source-id: 61e3e6c6b0f6b6b3d571bee02db2938ea9698c99

5 years agoUse at::zeros instead of torch::zeros in non-differentiable example (#15527)
Alexander Rodin [Wed, 26 Dec 2018 05:43:38 +0000 (21:43 -0800)]
Use at::zeros instead of torch::zeros in non-differentiable example (#15527)

Summary:
There was a typo in C++ docs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15527

Differential Revision: D13547858

Pulled By: soumith

fbshipit-source-id: 1f5250206ca6e13b1b1443869b1e1c837a756cb5

5 years agoFix the compare logic in function `overflows` for MSVC (#15499)
peter [Wed, 26 Dec 2018 05:43:22 +0000 (21:43 -0800)]
Fix the compare logic in function `overflows` for MSVC (#15499)

Summary:
Fixes https://github.com/pytorch/pytorch/issues/15497.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15499

Differential Revision: D13547835

Pulled By: soumith

fbshipit-source-id: a674da93bf905a0b81f0cc60449ccb97c2746926

5 years agoAllow converting char tensor to numpy; add [fi]info.min (#15046)
SsnL [Mon, 24 Dec 2018 17:08:50 +0000 (09:08 -0800)]
Allow converting char tensor to numpy; add [fi]info.min (#15046)

Summary:
https://github.com/pytorch/pytorch/pull/14710 with test fixed.

Also added `finfo.min` and `iinfo.min` to get castable tensors.

cc soumith
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15046

Reviewed By: soumith

Differential Revision: D13429388

Pulled By: SsnL

fbshipit-source-id: 9a08004419c83bc5ef51d03b6df3961a9f5dbf47

5 years agoPort replication_pad1d to ATen (#15507)
Lin Huang [Mon, 24 Dec 2018 14:29:34 +0000 (06:29 -0800)]
Port replication_pad1d to ATen (#15507)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15507

Pull Request resolved: https://github.com/pytorch/pytorch/pull/15485

port replication_pad1d

Reviewed By: ezyang

Differential Revision: D13531920

fbshipit-source-id: dcd64ebd2c24b7431996231b8d5addfb600b1072

5 years agoSupport stateful dataset (#15096)
Peter Goldsborough [Mon, 24 Dec 2018 14:23:32 +0000 (06:23 -0800)]
Support stateful dataset (#15096)

Summary:
Currently re-implements the dataloader for stateful datasets. Outstanding work:
- Refactor DataLoader and DataLoader2 to have common base classes and only differ in specifi pieces of logic,
- Figure out how to not duplicate the `MapDataset` logic for stateful vs. non-stateful
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15096

Differential Revision: D13522043

Pulled By: goldsborough

fbshipit-source-id: 08e461ca51783047f11facc4d27dfa2e4f1e4c2a

5 years agoput interactive prompt in bash (#15521)
Michael Suo [Mon, 24 Dec 2018 13:34:17 +0000 (05:34 -0800)]
put interactive prompt in bash (#15521)

Summary:
This makes compatibility with different versions of python a little bit simpler, and fixes a problem where stdin wasn't being read from the terminal properly in the prompt.

zdevito This should fix your EOF exception.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15521

Differential Revision: D13546358

Pulled By: suo

fbshipit-source-id: fb7551a86c888196831c046d9d9848e7ff05b925

5 years agoFix the iterator category for torch::data::Iterator (#15500)
peter [Mon, 24 Dec 2018 03:47:03 +0000 (19:47 -0800)]
Fix the iterator category for torch::data::Iterator (#15500)

Summary:
Try to fix https://github.com/pytorch/pytorch/issues/14410.
Additional info: From this [page](https://stackoverflow.com/questions/14062297/canonical-way-to-define-forward-output-iterator), If we change it into `input_iterator_tag`, it doesn't mean the `output_iterator_tag` is lost.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15500

Differential Revision: D13545773

Pulled By: soumith

fbshipit-source-id: 327bfb7be83d53e42925e0e391b2a4277e3a1b36

5 years agoPrecommit hook: just warn if no clang-tidy (#15514)
Michael Suo [Sun, 23 Dec 2018 22:35:41 +0000 (14:35 -0800)]
Precommit hook: just warn if no clang-tidy (#15514)

Summary:
The precommit hook shouldn't hard fail if there's no `clang-tidy`, just warn and omit the check.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15514

Differential Revision: D13545776

Pulled By: suo

fbshipit-source-id: 9bf3f8ee18703c6d1a39eb7776092fb5e120d2a1

5 years agoAdd torch.rot90 to torch.rst
Gao, Xiang [Sun, 23 Dec 2018 22:28:31 +0000 (14:28 -0800)]
Add torch.rot90 to torch.rst

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15512

Differential Revision: D13545775

Pulled By: soumith

fbshipit-source-id: 2a8896571745630cff4aaf3d5469ef646bdcddb4

5 years agofix parallelization detection for CPU foreach_reduced_elt (#15483)
Brennan Vincent [Sun, 23 Dec 2018 20:49:08 +0000 (12:49 -0800)]
fix parallelization detection for CPU foreach_reduced_elt (#15483)

Summary:
This does two things:

(1): revert #15114 , which is incorrect and actually just completely disables parallelization in this function (because `at::get_num_threads` returns `-1` unless it has been set explicitly)

(2): Fix our (FB-internal) failing tests that #15114 was intended to fix, by still working correctly in a setup where `#ifdef _OPENMP` is set and `omp_get_max_threads() > 1` , but `#pragma omp parallel` only launches one thread. I believe such an unusual situation only exists in certain unit tests within FB infra but we still need it to work.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15483

Differential Revision: D13538940

Pulled By: umanwizard

fbshipit-source-id: a3362c7ac7327ced350d127bb426f82c59e42732

5 years agoadd rowwise adagrad lp test (#15082)
Jongsoo Park [Sat, 22 Dec 2018 18:22:56 +0000 (10:22 -0800)]
add rowwise adagrad lp test (#15082)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15082

We didn't have unit test for low-precision rowwise adagrad

Reviewed By: chocjy

Differential Revision: D13300732

fbshipit-source-id: 46e7bdfc82c5a6855eeb6f653c0a96b0b3a20546

5 years agohandle empty inputs to SparseLengthsMean correctly (#15389)
Jongsoo Park [Sat, 22 Dec 2018 06:17:35 +0000 (22:17 -0800)]
handle empty inputs to SparseLengthsMean correctly (#15389)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15389

SparseLengthsMean was generating uninitialized data for empty inputs (lengths == 0). We should return zeros.
The unit tests were also not covering this special case which is fixed by this diff.

Reviewed By: salexspb

Differential Revision: D13515970

fbshipit-source-id: 3c35265638f64f13f0262cee930c94f8628005da

5 years agoAdd pthreadpool_create and pthreadpool_destroy (#15492)
Hao Lu [Sat, 22 Dec 2018 04:23:14 +0000 (20:23 -0800)]
Add pthreadpool_create and pthreadpool_destroy (#15492)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15492

Add pthreadpool_create and pthreadpool_destroy, which are used by NNPACK tests.

Reviewed By: Maratyszcza

Differential Revision: D13540997

fbshipit-source-id: 628c599df87b552ca1a3703854ec170243f04d2e

5 years agoMetadata for input/output formats in model file proto. (#15252)
Pritam Damania [Sat, 22 Dec 2018 01:34:51 +0000 (17:34 -0800)]
Metadata for input/output formats in model file proto. (#15252)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15252

We would like to extend the model file format to include strongly type, semantic information
about the model inputs and outputs.

The goal is for a user to be able to consider a model file like a function with
a well defined API describing what the inputs and outputs would be.

Reviewed By: dzhulgakov

Differential Revision: D13009915

fbshipit-source-id: 5df124a876ad03c05fbdaacae0eab659637734c1

5 years agoadd len to nativeResolver (#15488)
Zachary DeVito [Sat, 22 Dec 2018 00:44:19 +0000 (16:44 -0800)]
add len to nativeResolver (#15488)

Summary:
(otherwise len is not resolvable using torch::jit::compile)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15488

Differential Revision: D13539991

Pulled By: zdevito

fbshipit-source-id: 3ba85fa7b1adb163f9229c568f7997d22321903d

5 years agoRemove NoneGenerator
David Riazati [Sat, 22 Dec 2018 00:30:35 +0000 (16:30 -0800)]
Remove NoneGenerator

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15335

Differential Revision: D13540357

Pulled By: driazati

fbshipit-source-id: a289e5944b65872103f68faac74e18f10e7c6fff

5 years agoAdd self to Python printer reserved words (#15318)
David Riazati [Fri, 21 Dec 2018 23:59:29 +0000 (15:59 -0800)]
Add self to Python printer reserved words (#15318)

Summary:
This adds `self` to the list of reserved words and also sorts the lines and prevents the tracer from naming values 'self' (which happens in torch/tensor.py)

Fixes #15240
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15318

Differential Revision: D13540192

Pulled By: driazati

fbshipit-source-id: 46ae02e51b1b31d5c62110fa83ba258ea6bada27

5 years agoAD support for adaptive_avg_pool2d (#15459)
Ailing Zhang [Fri, 21 Dec 2018 23:32:44 +0000 (15:32 -0800)]
AD support for adaptive_avg_pool2d (#15459)

Summary:
This adds AD support for adaptive_avg_pool2d, which is necessary for resnet50 in pytorch/vision:master. cc: soumith asuhan dlibenzi

apaszke  I saw that autodiff bug you fixed in #15403 , as it doesn't prevent this PR from passing, so I'll leave it for your PR to fix it. :)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15459

Differential Revision: D13534732

Pulled By: ailzhang

fbshipit-source-id: 4e48b93e35d5ecfe7bd64b6a132a55b07843f206

5 years agoHandling nullptr case
Hao Lu [Fri, 21 Dec 2018 23:05:12 +0000 (15:05 -0800)]
Handling nullptr case

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15467

Reviewed By: Maratyszcza

Differential Revision: D13536504

fbshipit-source-id: ab46ff6bb4b6ce881c3e29d7e6a095ea62289db4

5 years agoRelax check on outputs (#15458)
Bram Wasti [Fri, 21 Dec 2018 22:11:26 +0000 (14:11 -0800)]
Relax check on outputs (#15458)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15458

many nets in the wild seem to have outputs that are never produced by the net.

Reviewed By: ZolotukhinM

Differential Revision: D13534185

fbshipit-source-id: 2b23b39c28404c53f68868f3bf6df53c5fea9eab

5 years agoallow non-final returns (#15463)
Zachary DeVito [Fri, 21 Dec 2018 21:46:12 +0000 (13:46 -0800)]
allow non-final returns (#15463)

Summary:
This PR allows a subclass of programs that have return statements that are not final in the graph.

`final_returns.h` contains the a comment describing how this is accomplished.
To minimize complexity in `compiler.cpp`, this pass is done as an AST-to-AST rewrite before the compiler runs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15463

Differential Revision: D13538962

Pulled By: zdevito

fbshipit-source-id: 67105ca873351825b4a364092ab1873779f3e462

5 years agoFixed trivial typos in Dropout2D and Dropout3D classes (#15200)
derek [Fri, 21 Dec 2018 19:54:57 +0000 (11:54 -0800)]
Fixed trivial typos in Dropout2D and Dropout3D classes (#15200)

Summary:
Fixed trivial typos in Dropout2D and Dropout3D classes

weiyangfb
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15200

Differential Revision: D13537888

Pulled By: ezyang

fbshipit-source-id: 8fb06027ca663a2e4bfa016af400698ae3c88ad1

5 years agoUpdating submodules
svcscm [Fri, 21 Dec 2018 19:44:29 +0000 (11:44 -0800)]
Updating submodules

Reviewed By: cdelahousse

fbshipit-source-id: 59d7a5b82fb78bc2d2285d0896e35c262512ffb9

5 years agoeq_fixes (#15475)
surgan12 [Fri, 21 Dec 2018 19:32:02 +0000 (11:32 -0800)]
eq_fixes (#15475)

Summary:
fixes #15464 .
cc : ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15475

Differential Revision: D13537812

Pulled By: ezyang

fbshipit-source-id: 127adf612ac8b3d3a64baa3d12a53daba7d3e4b8

5 years agoEnable running collect_env.py without building PyTorch (#15468)
vishwakftw [Fri, 21 Dec 2018 19:29:36 +0000 (11:29 -0800)]
Enable running collect_env.py without building PyTorch (#15468)

Summary: Closes #15346

Differential Revision: D13537873

Pulled By: ezyang

fbshipit-source-id: 7765ce4108dae9479d8900c0815cc2f174596a83

5 years agoBack out "[nomnigraph][executor] computeChains with nomnigraph" (#15451)
Bram Wasti [Fri, 21 Dec 2018 19:06:49 +0000 (11:06 -0800)]
Back out "[nomnigraph][executor] computeChains with nomnigraph" (#15451)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15451

Original commit changeset: ccd050bfead6

Reviewed By: ilia-cher

Differential Revision: D13533161

fbshipit-source-id: 1d0dcd54c2e3875aab015f3e996693e67a449b87

5 years agoDirect FBGEMM integraton into ATen (#13777)
James Reed [Fri, 21 Dec 2018 18:32:57 +0000 (10:32 -0800)]
Direct FBGEMM integraton into ATen (#13777)

Summary:
This PR implements infrastructure for post-processing a model to apply int8 quantization to its `nn.Linear` modules. Highlights of the implementation:

1) Inputs and outputs are `float` (quantized and packed internally), but the weight is quantized and packed ahead of time for efficiency. This implementation performs well in small-batch size GEMM calls. It should not be considered a general-purpose quantized GEMM kernel.
2) Weight packing is dependent on machine architecture (e.g. vector register width), so it is done just-in-time. Concretely, it is done on model load for the weights and it is done during operator execution for the input value.
3) Biases are unquantized
4) We fail loudly if we are attempting to run this on a machine that does not support FBGEMM. This is because we do not want a model's numerics to differ based on which machine it is run on. A model containing these FBGEMM ops *must* be run with FBGEMM

The API can be seen in the added test case. Highlights are:
1) `torch.jit.quantized.quantize_linear_modules` walks the module hierarchy of the passed-in Module and replaces all `nn.Linear` modules with a new `QuantizedLinear` module, which encapsulates the behavior described above.
2) `_pack()` and `_unpack()` script methods are present on `QuantizedLinear` modules. These methods should be called before serialization and after deserialization, respectively. This ensures that the weight matrix is properly packed for the running machine's architecture. Note that in the long term, we would like to move toward a more Pickle-style serialization technique, rather than having these explicit methods that mutate member values. This is blocked on being able to assign attributes in a ScriptMethod, among other things.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13777

Differential Revision: D13383276

Pulled By: jamesr66a

fbshipit-source-id: 00f29c9f34544add2b90107e3cf55a287802c344

5 years agoReplace getargspec with getfullargspec (#15396)
Ashwin Ramaswami [Fri, 21 Dec 2018 17:37:25 +0000 (09:37 -0800)]
Replace getargspec with getfullargspec (#15396)

Summary:
Replace `getargspec` with `getfullargspec` to resolve test warnings. Fixes #15344 .
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15396

Differential Revision: D13529548

Pulled By: zou3519

fbshipit-source-id: 50d3be92423a9ce89bc4895b67569663e1abbaa6

5 years agoThe benchmark binary support multiple batches in one run (#15443)
Fei Sun [Fri, 21 Dec 2018 16:39:05 +0000 (08:39 -0800)]
The benchmark binary support multiple batches in one run (#15443)

Summary:
It is sometimes beneficial to run multiple batches in one benchmark and check the aggregated results.

This PR enables this functionality.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15443

Reviewed By: llyfacebook

Differential Revision: D13531129

Pulled By: sf-wind

fbshipit-source-id: 553a762a5cbadf5a3d9fd6af767ae34899bc1aa2

5 years agoMove torch.logspace to ATen and parallelize on CPU.
Gregory Chanan [Fri, 21 Dec 2018 16:18:37 +0000 (08:18 -0800)]
Move torch.logspace to ATen and parallelize on CPU.

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15438

Reviewed By: ezyang

Differential Revision: D13529626

Pulled By: gchanan

fbshipit-source-id: 896e8afee3d6b5a706c4f5815b91ba6bd8af6672

5 years agoFix cudnn dropout (#15473)
Dmytro Dzhulgakov [Fri, 21 Dec 2018 16:13:15 +0000 (08:13 -0800)]
Fix cudnn dropout (#15473)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15473

Revert accidental changes introduced in D13335176

IntList is a range and copying it just copies pointers. Thus pointers would point either on deallocated memory or on the same memory causing equality always pass.

Reviewed By: ezyang

Differential Revision: D13537131

fbshipit-source-id: c97b3533be689bb4cdadd9e612f1284ac50e4bda

5 years agoformat specialized_segment_ops_test.py to prepare D13515970 (#15408)
Jongsoo Park [Fri, 21 Dec 2018 07:26:23 +0000 (23:26 -0800)]
format specialized_segment_ops_test.py to prepare D13515970 (#15408)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15408

Applied formatting to specialized_segment_ops_test.py to prepare D13515970

Reviewed By: salexspb

Differential Revision: D13520300

fbshipit-source-id: c3250b6abe8087c607f65ae60d1da61bd46c342b

5 years agoClean up onnxifi transformation code (#15453)
Yinghai Lu [Fri, 21 Dec 2018 06:04:09 +0000 (22:04 -0800)]
Clean up onnxifi transformation code (#15453)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15453

Just move things around to facilitate further development. No logic change.

Reviewed By: rdzhabarov

Differential Revision: D13533959

fbshipit-source-id: eebab1306939e802aacffb24a711d372fd67916c

5 years agoRecord Caffe2's current stream ID in c10_cuda. (#15174)
Edward Yang [Fri, 21 Dec 2018 05:51:25 +0000 (21:51 -0800)]
Record Caffe2's current stream ID in c10_cuda. (#15174)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15174

Previously, Caffe2 maintained a separate per-thread per-device
current logical CUDA stream ID.  In this PR, we switch Caffe2 over
to using c10::Stream to manage the current stream, and also
manage the allocation of cudaStream_t objects.

This results in a slight behavior change: previously, Caffe2
would have been willing to allocate an arbitrary number of
CUDA streams, depending on how high the logical stream IDs
went.  The c10::Stream pool has a fixed number of streams, once
you exceed it, it wraps around.

Reviewed By: dzhulgakov

Differential Revision: D13451550

fbshipit-source-id: da6cf33ee026932a2d873835f6e090f7b8a7d8dc

5 years agoAdd option to automatically handle unsorted variable-length sequences in RNNs (#15225)
Richard Zou [Fri, 21 Dec 2018 01:34:41 +0000 (17:34 -0800)]
Add option to automatically handle unsorted variable-length sequences in RNNs (#15225)

Summary:
Fixes #3584.

Motivation: manually sorting sequences, packing them, and then unsorting them
is something a lot of users have complained about doing, especially when we can
offer library support for them.

Overview: we internally sort sequences before packing them and store a list of
`unsorted_indices` that represent how to unsort the sequences inside
PackedSequence. The packing helper functions return PackedSequence with the
`permutation` field and the unpacking helper functions use it to unsort.

To implement this, the following changes were made:
- PackedSequence now keeps `sorted_indices` and `unsorted_indices`.
  These two can be thought of as permutations and are inverses of each other.
  `sorted_indices` is how the sequences were sorted; `unsorted_indices` is how
  to unsort the sequences.
- Added an `enforce_sorted` argument to pack_sequence and pack_padded_sequence
  that maintains the legacy behavior of error-ing out on unsorted-sequences.
  When `enforce_sorted=True`, these functions maintain their ONNX exportability.
- pack_sequence(sequences, enforce_sorted) takes in unsorted sequences.
- pack_padded_sequence can take in a padded tensor that represents padded,
  unsorted sequences.
- pad_packed_sequence unsorts the PackedSequence such that it is still the
  inverse operation of packed_padded_sequence.
- RNNs apply `sort_indices` to their input hidden state and apply
  `unsort_indices` to their output hidden state. This is to ensure that the
  hidden state batches correspond to the user's ordering of input sequences.

NOT BC-Breaking
- The default for pack_sequence and pack_padded_sequence is
  `enforce_sorted=True` to avoid breaking ONNX export. To use the new
  functionality, pass in `enforce_sorted=False`

Testing Plan
- Modified TestNN.test_pack_sequence, TestNN.test_packed_padded_sequence,
  and TestNN.test_variable_sequence (RNN test) to check the behavior
  of unsorted sequences, sorted sequences, and sorted sequences with
  enforce_sorted=True
- test/test_jit.py has a test to see if RNNs are exportable with
  enforce_sorted=True

cc colesbury
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15225

Reviewed By: soumith

Differential Revision: D13507138

Pulled By: zou3519

fbshipit-source-id: b871dccd6abefffca81bc4e3efef1873faa242ef

5 years agoChange default value of unique to 'sorted=True'
WeihuangXu [Fri, 21 Dec 2018 01:04:14 +0000 (17:04 -0800)]
Change default value of unique to 'sorted=True'

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15379

Differential Revision: D13531287

Pulled By: ezyang

fbshipit-source-id: 1512da7d660dc413688d99264e6434897c3ac78c

5 years agoadd denormal options (ftz and daz)
Jongsoo Park [Fri, 21 Dec 2018 01:01:53 +0000 (17:01 -0800)]
add denormal options (ftz and daz)

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15423

Reviewed By: yinghai

Differential Revision: D13526340

fbshipit-source-id: de2ecc717b4f778f33a8bf940ed144dbb230c7a8

5 years agocollect_env fix (#15447)
surgan12 [Fri, 21 Dec 2018 00:53:49 +0000 (16:53 -0800)]
collect_env fix (#15447)

Summary:
fixes #15214
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15447

Differential Revision: D13531523

Pulled By: ezyang

fbshipit-source-id: 8f24f5ae9f3e78f6c5c9ee702ba14faca7aa297a

5 years agoRemove unused field in jit script module deserializer (#15439)
Lu Fang [Fri, 21 Dec 2018 00:14:16 +0000 (16:14 -0800)]
Remove unused field in jit script module deserializer (#15439)

Summary:
A little bit clean up.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15439

Reviewed By: zrphercule

Differential Revision: D13532015

Pulled By: houseroad

fbshipit-source-id: 2fb1e01fc28549c7e78af6c65ee68339950bc7da

5 years agoRevert D13494873: [pytorch][PR] Fixing ONNX export of logical ops to have correct...
Edward Yang [Thu, 20 Dec 2018 23:44:09 +0000 (15:44 -0800)]
Revert D13494873: [pytorch][PR] Fixing ONNX export of logical ops to have correct output datatype

Differential Revision:
D13494873

Original commit changeset: 069d2f956a5a

fbshipit-source-id: 80ef10b2eb623a63da51dc2e4874f2ee446f426d

5 years agoFix ASAN div by zero error in rotated GenerateProposals op (#15415)
Viswanath Sivakumar [Thu, 20 Dec 2018 23:33:44 +0000 (15:33 -0800)]
Fix ASAN div by zero error in rotated GenerateProposals op (#15415)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15415

Was introduced in D13429770

Reviewed By: SuperIRabbit

Differential Revision: D13524114

fbshipit-source-id: a890eb3b97c24952c361155d1432a801499f4ddd

5 years agoTensor construction codemod(ResizeLike) - 7/7 (#15087)
Jerry Zhang [Thu, 20 Dec 2018 23:28:12 +0000 (15:28 -0800)]
Tensor construction codemod(ResizeLike) - 7/7 (#15087)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15087

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

Reviewed By: ezyang

Differential Revision: D13419765

fbshipit-source-id: 34d695309a66723281429610a12544598c507d74

5 years agoallow numpy-like boolean-list indexing in pytorch (#14932)
rory [Thu, 20 Dec 2018 23:18:39 +0000 (15:18 -0800)]
allow numpy-like boolean-list indexing in pytorch (#14932)

Summary:
Suggested fix to issue #6773, the fix allows numpy-like boolean-list indexing in pytorch
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14932

Differential Revision: D13398795

Pulled By: ezyang

fbshipit-source-id: 67f8daf9829db2550ff76d2bde673be6dd2708cd

5 years agoDoc improvement on DDP (#15440)
Teng Li [Thu, 20 Dec 2018 22:46:01 +0000 (14:46 -0800)]
Doc improvement on DDP (#15440)

Summary:
I noticed that some users don't even know we have this support. Adding into the doc
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15440

Differential Revision: D13531045

Pulled By: teng-li

fbshipit-source-id: 9757c400c0010608758c754df04e603b36035a10

5 years agoFix type annotation error. (#15448)
Edward Yang [Thu, 20 Dec 2018 22:26:23 +0000 (14:26 -0800)]
Fix type annotation error. (#15448)

Summary:
According to mypy, the trailing -> None is mandatory.

Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15448

Differential Revision: D13532179

Pulled By: ezyang

fbshipit-source-id: e8972f8c9ada4657c518cd7bcd46e489ab8ddf5f

5 years agoAdd launch bounds needed for ROCm 2.0 (#15400)
Johannes M Dieterich [Thu, 20 Dec 2018 22:26:14 +0000 (14:26 -0800)]
Add launch bounds needed for ROCm 2.0 (#15400)

Summary:
ROCm 2.0's compiler requires launch_bounds annotations if flat work group sizes are larger than the default of 256.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15400

Differential Revision: D13531239

Pulled By: ezyang

fbshipit-source-id: c0b40600a8c332823da6c7113c644d8dba424a9c