Helmut [Mon, 28 Jan 2019 19:25:33 +0000 (11:25 -0800)]
Fix compiler error in swapBytes64 for rare architectures (#16418)
Summary:
swapBytes64 used to use SwapByteOrder_32 and value, both of which dont exist. This commit rewrites that part from scratch.
This happened on Debugbuild on Microsoft compiler. For that case " && !defined(_DEBUG)" is also removed, because _byteswap_uint64 works fine in debug mode (if it is necessary it should me commented why).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16418
Differential Revision:
D13843306
Pulled By: ezyang
fbshipit-source-id:
dde1c7baeccec3aaa750d4b7200b3f4ccb4a00cb
Junjie Bai [Mon, 28 Jan 2019 19:10:18 +0000 (11:10 -0800)]
Fix lint errors introduced in pytorch/pytorch@ceece5d (#16454)
Summary:
ifedan
```
./test/common_utils.py:748:1: E302 expected 2 blank lines, found 1
./test/test_torch.py:1235:5: E303 too many blank lines (2)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16454
Differential Revision:
D13844905
Pulled By: bddppq
fbshipit-source-id:
3dc7c740d86310a8efc9864d7c7798fda8257a21
Syed Tousif Ahmed [Mon, 28 Jan 2019 18:20:47 +0000 (10:20 -0800)]
Report the slowest 10 tests when using pytest (#16423)
Summary:
This flag is useful in identifying if a test is taking way too long like the ones in the following snippet when running the test suite with pytest. https://github.com/pytorch/pytorch/blob/
9757ad35b0b56cf955f294e751de9b437f9bb4ff/test/common_utils.py#L814-L835
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16423
Differential Revision:
D13843507
Pulled By: ezyang
fbshipit-source-id:
643e1766a85905b3b112ea5ca562135a17896a72
Xiaomeng Yang [Mon, 28 Jan 2019 17:26:41 +0000 (09:26 -0800)]
Optimize SpatialBNOp on GPU (#16395)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16395
Optimize SpatialBNOp on GPU
i-am-not-moving-c2-to-c10
Reviewed By: houseroad
Differential Revision:
D13829833
fbshipit-source-id:
04d2a63e8e9830c4c39a91cf87fcd7aa765dc55f
Igor Fedan [Mon, 28 Jan 2019 17:14:07 +0000 (09:14 -0800)]
CPU implementation of torch.cdist (#16168)
Summary:
cdist is used for calculating distances between collections of observations.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16168
Differential Revision:
D13739147
Pulled By: ifedan
fbshipit-source-id:
9419c2c166891ac7db40672c72f17848f0b446f9
Brennan Vincent [Mon, 28 Jan 2019 16:47:35 +0000 (08:47 -0800)]
Don't initialize a new `std::vector` in a loop. (#15850)
Summary:
Before this diff, we execute `std::vector<optional<acc_t>> buffer((unsigned)max_threads, optional<acc_t> {});` in every iteration of `foreach_reduced_elt`. Change the code to only execute that line if we need it; i.e., we are actually about to parallelize.
This overhead is quite significant when we are doing a lot of small reductions in single-threaded code.
```
x=torch.randn((1024,10,1024),dtype=torch.float64)
torch.set_num_threads(1)
%timeit x.std(1)
```
Before (with #15845 applied): 708.25 ms
After: 508 ms
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15850
Differential Revision:
D13612960
Pulled By: umanwizard
fbshipit-source-id:
f5e61abfe0027775c97ed81ac09c997fbee741df
Edward Yang [Mon, 28 Jan 2019 15:37:43 +0000 (07:37 -0800)]
More documentation on caffe2::Operator
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16371
Reviewed By: dzhulgakov
Differential Revision:
D13820472
fbshipit-source-id:
efccea0e92c86d30ec2bdda50eb9aab8a3a1504d
rotuna [Mon, 28 Jan 2019 00:26:47 +0000 (16:26 -0800)]
Better error message when creating a module instance in jit.script (#16416)
Summary:
Made the change requested in #15555
PR was failing build due to a time out error while getting packages using pip.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16416
Differential Revision:
D13833873
Pulled By: soumith
fbshipit-source-id:
e2200e9e8015558fcd359dfa3d025b25802d62b5
peter [Sun, 27 Jan 2019 22:59:34 +0000 (14:59 -0800)]
Fix issues on Windows brought by #16289 (#16412)
Summary:
This one needs to be merged ASAP because the CUDA build for Windows is skipped at this time.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16412
Differential Revision:
D13833889
Pulled By: soumith
fbshipit-source-id:
95a401a01fb0f9c1045df0bfd72d8206b8a6f3fd
Gemfield [Sun, 27 Jan 2019 22:13:46 +0000 (14:13 -0800)]
Fix a typo in Parallel.h (#16419)
Summary:
Fix a typo in Parallel.h.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16419
Differential Revision:
D13833705
Pulled By: soumith
fbshipit-source-id:
824ebe753e028fc8e2b5d7a51fdba98a365fd29a
peterjc123 [Sun, 27 Jan 2019 20:26:24 +0000 (12:26 -0800)]
Don't install PDB for Windows static build of caffe2_observers (#16420)
Summary:
Fixes #16292.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16420
Differential Revision:
D13833704
Pulled By: soumith
fbshipit-source-id:
482ad6ce103bed7206e924e8c82454fbb1bfac42
SsnL [Sun, 27 Jan 2019 20:08:09 +0000 (12:08 -0800)]
Fix slogdet sign requiring grad when input requires grad (#16337)
Summary:
The real fix for https://github.com/pytorch/pytorch/issues/15605.
This is sort of BC breaking because now
```py
In [1]: import torch
In [2]: a = torch.randn(3, 3, requires_grad=True)
In [3]: a.slogdet()
Out[3]: (tensor(1.), tensor(0.1356, grad_fn=<SlogdetBackward>))
In [4]: a.slogdet()[0].requires_grad
Out[4]: False
```
while before this patch ` a.slogdet()[0]` requires grad with `grad_fn=<SlogdetBackward>`. But any use of backproping through this value will meet the error in #15605 so I don't think this is a problem.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16337
Differential Revision:
D13832644
Pulled By: soumith
fbshipit-source-id:
f96c477e99edcbdbd966888e5c5ea7fd058429a8
Zachary DeVito [Sun, 27 Jan 2019 09:24:58 +0000 (01:24 -0800)]
CI Fix: restore MAX_JOBS variable (#16415)
Summary:
Restores a CI workaround (https://github.com/pytorch/pytorch/pull/7361) that got dropped with build_pytorch_libs.sh.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16415
Differential Revision:
D13833092
Pulled By: zdevito
fbshipit-source-id:
f78b60cafd8da945790dba28de373b8faf46e9f5
Samuel Fadel [Sun, 27 Jan 2019 01:58:31 +0000 (17:58 -0800)]
Update einsum documentation. (#16323)
Summary:
The documentation stated that operands to einsum should be a list of Tensors, not individual arguments. The function, however, now accepts individual arguments for each Tensor operand *and* a single argument consisting of a list of Tensors. The documentation was updated to reflect this change.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16323
Differential Revision:
D13832647
Pulled By: soumith
fbshipit-source-id:
c01c2b350f47674d3170337f493b0ee2ea381b3f
James Reed [Sun, 27 Jan 2019 01:39:34 +0000 (17:39 -0800)]
Fix flake8 warnings/errors in test_jit.py (#16409)
Summary:
These were really annoying to see in the phabricator UI when trying to land PRs that touched test_jit.py, so this fixes them.
One remaining item is the T484 error. Locally, flake8 still chokes on that line even though I put the noqa comment there (and tried varying whitespaces around it etc). Not sure why it still persists...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16409
Differential Revision:
D13832658
Pulled By: jamesr66a
fbshipit-source-id:
46356ba6444ae5ee1a141c28489bdcc7c99e39c0
James Reed [Sat, 26 Jan 2019 22:38:12 +0000 (14:38 -0800)]
Trace fork and join calls
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16232
Differential Revision:
D13772974
Pulled By: jamesr66a
fbshipit-source-id:
b2db370271809e26d3301f8cc98eec567db5e62b
vishwakftw [Sat, 26 Jan 2019 19:14:19 +0000 (11:14 -0800)]
Switch to CUDA implementation if batch size >= 65536 for affine_grid (#16403)
Summary:
Changelog:
- Append a condition that switches to the native CUDA implementation for affine_grid
Fixes #16365
Differential Revision:
D13832192
Pulled By: soumith
fbshipit-source-id:
3f484e6673d71e3ba7627b170cb8f1611e12b9b2
SsnL [Sat, 26 Jan 2019 17:42:48 +0000 (09:42 -0800)]
gitignore gdb history
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16404
Differential Revision:
D13832191
Pulled By: soumith
fbshipit-source-id:
ab23d1ad72c041ec2d9616c273bbf399e0feb10d
Juan Miguel Pino [Sat, 26 Jan 2019 06:49:00 +0000 (22:49 -0800)]
Revert
D13821061: [redo][c10] layernorm example
Differential Revision:
D13821061
Original commit changeset:
82f0dade0145
fbshipit-source-id:
e5b0b1bab0c9e731ae04add35e9a6c91656dd178
Jerry Zhang [Sat, 26 Jan 2019 00:59:07 +0000 (16:59 -0800)]
trying to fix testX (#16370)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16370
passed locally but seems testX has some problem
Reviewed By: ezyang
Differential Revision:
D13820250
fbshipit-source-id:
e4ad9d1ec99508867d4ead46753a7fb7019c50bd
Bram Wasti [Sat, 26 Jan 2019 00:45:35 +0000 (16:45 -0800)]
layernorm example (#16374)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16374
this fixes the original attempt in OSS (adds to CMake and python build files)
Reviewed By: smessmer
Differential Revision:
D13821061
fbshipit-source-id:
82f0dade0145fd04bdf8e3cb3954b5790e918162
Bram Wasti [Sat, 26 Jan 2019 00:45:34 +0000 (16:45 -0800)]
plug caffe2 into jit" (#16388)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16388
previous diff broke master -- this refactors out the custom_operator.cpp file into a separate header + cpp pair (caffe2_operator.{h,cpp})
Reviewed By: smessmer
Differential Revision:
D13823550
fbshipit-source-id:
00e005e650336132d05aef97c1f0e5242ccad5ba
Junjie Bai [Sat, 26 Jan 2019 00:06:59 +0000 (16:06 -0800)]
Enable centos pytorch rocm CI
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14879
Differential Revision:
D13821534
Pulled By: bddppq
fbshipit-source-id:
45151b880992f1efa83e29c4985a723374575506
Zachary DeVito [Fri, 25 Jan 2019 23:57:09 +0000 (15:57 -0800)]
Remove bash from build (#16289)
Summary:
This commit removes the dependency on `build_pytorch_libs.sh` by moving the remaining functionality that is not expressible in cmake into python. Removing the indirection through bash also removes over 300 lines of environment munging code that is incredibly hard to understand because it passes a lot of secret parameters through `os.env`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16289
Reviewed By: ezyang
Differential Revision:
D13821662
Pulled By: zdevito
fbshipit-source-id:
d658d26925e3b1169ac1e3d44a159cf8a1f0d9b1
Jerry Zhang [Fri, 25 Jan 2019 23:32:45 +0000 (15:32 -0800)]
Remove caffe2::ShareData (#16139)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16139
Original commit changeset:
4b15a4c62995
Reviewed By: dzhulgakov
Differential Revision:
D13677464
fbshipit-source-id:
1a644a88fac02b44feebac48ccc01bc72cc47edb
Jesse Hellemn [Fri, 25 Jan 2019 23:11:33 +0000 (15:11 -0800)]
Trying a fix to anaconda logins on nightlies
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16387
Differential Revision:
D13826227
Pulled By: pjh5
fbshipit-source-id:
769a53e40a4912879faf9716a80c0e0c86acdbf8
Elias Ellison [Fri, 25 Jan 2019 23:02:30 +0000 (15:02 -0800)]
Update Documentation for Optionals (#16380)
Summary:
Now that https://github.com/pytorch/pytorch/pull/15587 has landed, updating docs.
Will close https://github.com/pytorch/pytorch/issues/15278
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16380
Differential Revision:
D13825221
Pulled By: eellison
fbshipit-source-id:
c5a7a7fbb40ba7be46a80760862468f2c9967169
Zachary DeVito [Fri, 25 Jan 2019 20:20:29 +0000 (12:20 -0800)]
Revert
D13740752: [c10] plug caffe2 into jit
Differential Revision:
D13740752
Original commit changeset:
2d9383574d42
fbshipit-source-id:
e9ff217a438720423340a10af7fa263b33f2ae24
Gu, Jinghui [Fri, 25 Jan 2019 19:00:32 +0000 (11:00 -0800)]
Impl Shape op for mkldnn (#15266)
Summary:
Impl Shape op for mkldnn
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15266
Differential Revision:
D13804558
Pulled By: yinghai
fbshipit-source-id:
8a35f608c23973d7a15c3d645aee4059eb55f245
Bram Wasti [Fri, 25 Jan 2019 18:10:07 +0000 (10:10 -0800)]
Back out "[c10] layernorm example"
Summary: Original commit changeset:
87240ca7f48d
Reviewed By: bddppq
Differential Revision:
D13816657
fbshipit-source-id:
bafcf0779d811c7e4a134cfb323a89352fa8c180
Ailing Zhang [Fri, 25 Jan 2019 16:35:55 +0000 (08:35 -0800)]
Add xla test in CI (#15978)
Summary:
Adding xla CPU tests in our CI.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15978
Differential Revision:
D13816344
Pulled By: ailzhang
fbshipit-source-id:
f74c52e846976ea4ac439313847908a0e99d05eb
Edward Yang [Fri, 25 Jan 2019 15:48:00 +0000 (07:48 -0800)]
Delete Tensor::swap(), replace with pointer swap (#12730)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12730
i-am-not-moving-c2-to-c10
Reviewed By: smessmer
Differential Revision:
D10415430
fbshipit-source-id:
8a2ce8611c5fa77bbbd73fb6788c1baa3b370f07
SsnL [Fri, 25 Jan 2019 15:24:18 +0000 (07:24 -0800)]
Make test_proper_exit more robust (#16249)
Summary:
1. Improve error message for better debugging info
2. Increase timeout
3. Also apply the windows worker failure detection mechanism on non-Windows platforms, for better robustness
Attempt to fix #14501
cc ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16249
Differential Revision:
D13784702
Pulled By: ezyang
fbshipit-source-id:
09a7cff83ab9edce561ed69f9fb555ab35d1275f
Si Chen [Fri, 25 Jan 2019 15:23:06 +0000 (07:23 -0800)]
fix contbuild (#16362)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16362
https://our.intern.facebook.com/intern/testinfra/diagnostics/
281475065177800.
844424930381786.
1548397180/
Reviewed By: ezyang
Differential Revision:
D13816639
fbshipit-source-id:
024117233f6d3bc6244013ca2ee1aea065560212
Xiaomeng Yang [Fri, 25 Jan 2019 08:54:09 +0000 (00:54 -0800)]
Minor change of group_norm_gradient on GPU (#16307)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16307
Minor change of group_norm_gradient on GPU
Reviewed By: houseroad
Differential Revision:
D13800613
fbshipit-source-id:
9e55f93b1e322efe3fc2d684b9c47c3dbb7a0f48
Junjie Bai [Fri, 25 Jan 2019 07:50:21 +0000 (23:50 -0800)]
Revert
D13551909: [fbcode] logdevice for generic feature type
Differential Revision:
D13551909
Original commit changeset:
807830c50bee
fbshipit-source-id:
48cacf4ec1765253a9be9d78f4b28cc48330be59
Qin Huang [Fri, 25 Jan 2019 07:21:25 +0000 (23:21 -0800)]
logdevice for generic feature type (#16191)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16191
logdevice related modifications for generic feature type
we directly convert the generic feature structures to json strings, which corresponds to the column input in offline and dper
Reviewed By: itomatik
Differential Revision:
D13551909
fbshipit-source-id:
807830c50bee569de202530bc3700374757793a2
Bram Wasti [Fri, 25 Jan 2019 05:46:50 +0000 (21:46 -0800)]
layernorm example (#16350)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16350
Example usage of the new caffe2 integration
Reviewed By: smessmer
Differential Revision:
D13408546
fbshipit-source-id:
87240ca7f48d653a70241d243aa0eb25efa67611
Bram Wasti [Fri, 25 Jan 2019 05:46:50 +0000 (21:46 -0800)]
plug caffe2 into jit (#16331)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16331
Temporary measure to enable caffe2 ops in pytorch
Reviewed By: smessmer
Differential Revision:
D13740752
fbshipit-source-id:
2d9383574d42ce84ee471aba32eeb4f5a0cc7a4c
Bram Wasti [Fri, 25 Jan 2019 05:46:50 +0000 (21:46 -0800)]
Add RunOperator for using FunctionSchema registered ops easily in caffe2 (#16173)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16173
Helper to make it easy to run ops in caffe2
Reviewed By: smessmer
Differential Revision:
D13468240
fbshipit-source-id:
2276c7870af6dcdf829957f005fd16ac1ef319b5
Bram Wasti [Fri, 25 Jan 2019 05:46:50 +0000 (21:46 -0800)]
Add correct Input() shim to caffe2 operator impl (#16048)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16048
This enables full shimming of the operator (previously it was only
Output() shimmed).
Reviewed By: smessmer
Differential Revision:
D13468241
fbshipit-source-id:
c853b775ab5cdcd968f4a6cc4766e91c3c6b1c45
Shen Li [Fri, 25 Jan 2019 01:11:26 +0000 (17:11 -0800)]
Relax lower bound for nogil timing test to avoid false alarm (#16259)
Summary:
fixes #16250, #16271
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16259
Differential Revision:
D13784505
Pulled By: mrshenli
fbshipit-source-id:
0b7ad98cd3c018b9907d70158de3abc3c4cb57ef
Mikhail Zolotukhin [Fri, 25 Jan 2019 00:34:46 +0000 (16:34 -0800)]
Code-style fixes. (#16342)
Summary:
Some cleanups in ir.{h,cpp}. I plan to continue cleaning it up, so this is a first step.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16342
Differential Revision:
D13808897
Pulled By: ZolotukhinM
fbshipit-source-id:
2dedb414576c3efbf8e36434145d7f14a66b1ee7
Jongsoo Park [Fri, 25 Jan 2019 00:32:24 +0000 (16:32 -0800)]
disable testing group conv with EIGEN engine (#16335)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16335
group conv is not implemented with EIGEN engine so this diff disables related tests
Reviewed By: jamesr66a
Differential Revision:
D13807204
fbshipit-source-id:
41f6de43da40882f57e64474520e185733caefb7
Elias Ellison [Thu, 24 Jan 2019 23:41:50 +0000 (15:41 -0800)]
Remove unneeded manual unwrap optionals (#16245)
Summary:
Remove calls to torch.jit._unwrap_optional that are no longer needed.
The remaining instances would require control flow logic for exceptions.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16245
Differential Revision:
D13804292
Pulled By: eellison
fbshipit-source-id:
08c5cbe4b956519be2333de5cf4e202488aff626
Yan Shang [Thu, 24 Jan 2019 23:22:01 +0000 (15:22 -0800)]
fix buildindexop (#16341)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16341
as in the title
Reviewed By: intermilan
Differential Revision:
D13808679
fbshipit-source-id:
0d12d3253f380bec66bc9be899be565861b8163a
Hoa Dinh [Thu, 24 Jan 2019 23:20:09 +0000 (15:20 -0800)]
Revert
D13747581: Optimize SpatialBN on GPU
Differential Revision:
D13747581
Original commit changeset:
48a885a240ef
fbshipit-source-id:
58cec6023843d7459865eb80c9db8dac463cb96c
Jerry Zhang [Thu, 24 Jan 2019 23:01:47 +0000 (15:01 -0800)]
Add Test for ReinitializeTensor (#16338)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16338
att
Reviewed By: ezyang
Differential Revision:
D13806760
fbshipit-source-id:
322b9b7d314aeb0194f52b803ca35c0cb8efcdec
Will Feng [Thu, 24 Jan 2019 22:29:06 +0000 (14:29 -0800)]
Add thread-local guard: at::AutoNonVariableTypeMode (#15939)
Summary:
This PR adds thread-local guard (`at::AutoNonVariableTypeMode`) to make sure that in VariableType.cpp the operations on baseType still dispatch to non-Variable type, even if the parameters will become Variables after the Tensor/Variable merge. We achieve this by making `legacyTensorType()` and `getType()` check the `at::AutoNonVariableTypeMode` guard to decide whether to return non-Variable type for a variable.
This is part of the VariableImpl/TensorImpl merge work: https://github.com/pytorch/pytorch/issues/13638.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15939
Reviewed By: ezyang
Differential Revision:
D13640980
Pulled By: yf225
fbshipit-source-id:
d12c2543822958558d7d70d36c50999a5eb8783f
Jongsoo Park [Thu, 24 Jan 2019 22:09:11 +0000 (14:09 -0800)]
reduce parameter space of test_1x1_conv to avoid timeout (#16223)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16223
As title says
Reviewed By: jamesr66a
Differential Revision:
D13758202
fbshipit-source-id:
3cdffb80a5dad53b29e65e8eb0ae128edba70dbb
Sidney Zhang [Thu, 24 Jan 2019 21:07:35 +0000 (13:07 -0800)]
Update docs to include variable annotation example (#16324)
Summary:
Relates to this issue https://github.com/pytorch/pytorch/issues/16288
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16324
Reviewed By: ezyang
Differential Revision:
D13805412
Pulled By: suo
fbshipit-source-id:
8b80f988262da2c717452a71142327bbc23d1b8f
Edward Yang [Thu, 24 Jan 2019 20:00:34 +0000 (12:00 -0800)]
Delete duplicate copy of THCCachingAllocator. (#16226)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16226
Now that the caching allocator is moved to c10_cuda, we can
delete the duplicate copy from Caffe2.
Reviewed By: dzhulgakov, smessmer
Differential Revision:
D13762540
fbshipit-source-id:
03f1ebf7f11c68c19aa0d66110156fe228da6138
Edward Yang [Thu, 24 Jan 2019 20:00:34 +0000 (12:00 -0800)]
Move THCCachingAllocator to c10_cuda. (#16119)
Summary:
Some renaming and renamespacing also took place. I was originally planning not to do anything, but it turns out that it was easier to make HIPify work by using a namespace CUDACachingAllocator:: rather than THCCachingAllocator_, since :: is a word boundary but _ is not.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16119
Reviewed By: smessmer
Differential Revision:
D13718768
fbshipit-source-id:
884a481d99027fd3e34471c020f826aa12225656
Edward Yang [Thu, 24 Jan 2019 20:00:34 +0000 (12:00 -0800)]
Remove unnecessary includes and headers from THCCachingAllocator, move to at::cuda:: namespace (#16117)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16117
This means I can move it to c10_cuda with minimal fuss.
Reviewed By: smessmer
Differential Revision:
D13717836
fbshipit-source-id:
a94c7dc649af64542480fc1c226b289588886c00
Mikhail Zolotukhin [Thu, 24 Jan 2019 19:05:07 +0000 (11:05 -0800)]
Directly include headers from ATen.
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16287
Differential Revision:
D13792949
Pulled By: ZolotukhinM
fbshipit-source-id:
d627d8dc469df048063c70d0b5b8d33fede809a3
Richard Zou [Thu, 24 Jan 2019 19:01:57 +0000 (11:01 -0800)]
Refactor the docs build workflow (#16265)
Summary:
In preparation for setting up a doc build job for stable docs, I wanted
to refactor the workflow so that future changes will be easier.
This PR the following changes:
- Refactor the doc push script into a reusable command
- Add command line options for the doc push script.
These don't matter too much for now but will be useful
for setting up future jobs for building different versions of the
docs.
- Instead of checking out pytorch/pytorch:master, we re-use the pytorch
installation inside the docker image.
- Change the sed in the script to a perl command. sed is annoyingly
different across platforms; the perl command is more stable
- Run the script in dry run mode (without pushing the doc build)
whenever a PR is opened. This lets us test changes to the doc build workflow.
Test Plan
- I tested the doc build script locally with my own credentials and it
worked fine.
- Wait for the pytorch_doc_push CI.
- After merging this PR, keep an eye on the pytorch_doc_push CI status.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16265
Differential Revision:
D13803511
Pulled By: zou3519
fbshipit-source-id:
4564bca3e74d490f89a1d1da9fb8b98eb44bdbb1
Owen Anderson [Thu, 24 Jan 2019 18:40:38 +0000 (10:40 -0800)]
Save a little bit of work in constant pooling by not moving nodes that will get deleted.
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16161
Differential Revision:
D13791247
Pulled By: resistor
fbshipit-source-id:
2a5a4f98309509b4ba875373ee57e6f63c75a4fd
Gregory Chanan [Thu, 24 Jan 2019 15:36:54 +0000 (07:36 -0800)]
Handle non-contiguous inputs with mkldnn convolution. (#16300)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/16018.
Backwards appears to be fine because the derivative is written in terms of mkldnn_convolution itself.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16300
Differential Revision:
D13797776
Pulled By: gchanan
fbshipit-source-id:
68a990b8a3c186412a99d176931314806c9ed7bf
Xiaomeng Yang [Thu, 24 Jan 2019 10:50:35 +0000 (02:50 -0800)]
Optimize SpatialBN on GPU (#16202)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16202
Optimize SpatialBN on GPU
Reviewed By: houseroad
Differential Revision:
D13747581
fbshipit-source-id:
48a885a240ef2a325235e8f89ebbe50e7c780c84
Xiaomeng Yang [Thu, 24 Jan 2019 07:55:06 +0000 (23:55 -0800)]
optimize group_norm (#16216)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16216
Optimize GroupNormOp
Reviewed By: houseroad
Differential Revision:
D13754145
fbshipit-source-id:
650f64c81486c6c9d276f2e3325392d5838751ba
Lu Fang [Thu, 24 Jan 2019 05:32:57 +0000 (21:32 -0800)]
Fix the tensor deserialization problem of jit script module on CUDA (#16279)
Summary:
Now we create a temporary tensor for the whole record.
Fix https://github.com/pytorch/pytorch/issues/15271
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16279
Reviewed By: BIT-silence
Differential Revision:
D13791442
Pulled By: houseroad
fbshipit-source-id:
6f52ca09627fb684f74121357cc42e4adadec36a
Erik Brinkman [Thu, 24 Jan 2019 03:37:29 +0000 (19:37 -0800)]
Small fixes for pdist (#16210)
Summary:
pdist was recently patched to remove buggy batch support and fix issues
with large tensors. This fixed missed a few spots, and didn't handle a
few recommendations that this commit addresses.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16210
Differential Revision:
D13791914
Pulled By: gchanan
fbshipit-source-id:
0595841be1b298f7268fd4c02a6628acfec918f2
Jerry Zhang [Thu, 24 Jan 2019 03:24:38 +0000 (19:24 -0800)]
Fix comparison in ReinitializeTensor (#16294)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16294
In `ReinitializeTensor`, we compare `tensor->GetDevice()` and `options.device()`, but in the callsite, we actually just provide an option with `device_type`, which means the `device_id` will always be default(-1) for `options`, but for tensor, although it is passed a `device` with default `device_id`, when we allocate the data, the `device` of the `tensor` is the `device` of `Storage`, which is the `device` of underlying `DataPtr`, which is the same as the `device` of the `Context` of the operator, which has a non-default `device_id`.
Therefore everytime we do `ReinitializeTensor`, we'll find the `device` does not match, and after the `ReinitializeTensor` call, the `device` still does not match. That's why everytime we'll allocate a new Tensor and cause perf regressions for ops that uses `ReinitializeTensor` on multiple GPUs.
Reviewed By: BIT-silence
Differential Revision:
D13795635
fbshipit-source-id:
24d6afa1a0196a32eb0134ee08b4280244cdb0c3
Benny Chen [Thu, 24 Jan 2019 03:02:14 +0000 (19:02 -0800)]
Fix issues under caffe round 1
Summary: Some automation to fix uninitialized members for caffe2 code. Ran canary to make sure I don't have any regression in prod, but not sure how to test comprehensively for caffe2
Reviewed By: ezyang
Differential Revision:
D13776185
fbshipit-source-id:
fb2a479971cc0276d8784be1c44f01252410bd24
David Riazati [Thu, 24 Jan 2019 02:11:04 +0000 (18:11 -0800)]
Add support for overloaded functions (#15556)
Summary:
This PR adds support for overloaded functions as a step toward adding rnn modules to the JIT standard library.
Possible overloads must be manually specified, and when resolving the overload it chooses by the first one that passes the schema matching logic. The structure is very similar to boolean dispatch in #14425. The overload will only work on weak modules.
In order to avoid supporting overloaded methods in Python to match the JIT execution, the current setup offloads that work to the user. In the test added in `test_jit.py`, two methods are used to overload the `forward` method. In order to call `forward` outside the JIT, a Python-only `forward` that does the right argument type switching must also be provided.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15556
Differential Revision:
D13576348
Pulled By: driazati
fbshipit-source-id:
7d3bdd4ee5a6088cc20c92f26a696d1ee5b9204b
Elias Ellison [Thu, 24 Jan 2019 01:47:29 +0000 (17:47 -0800)]
Constant propagation changes (#16244)
Summary:
- remove loop node that is guaranteed not to execute
- remove extra loop outputs that are no longer needed
- if we are inlining an if node, only run constant propagation on the block that will execute
- remove the recurse argument since we only expose the Graph Constant Propagation and it's not used
This also includes a few extra hooks to python_ir that I think make it a little be easier to test graph conditions from python.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16244
Differential Revision:
D13791635
Pulled By: eellison
fbshipit-source-id:
d16351fffcfc8013b02015db200f8fde002e0577
nlml [Thu, 24 Jan 2019 00:02:55 +0000 (16:02 -0800)]
raise exception if try jit.load non-existent file (#16270)
Summary:
addresses https://github.com/pytorch/pytorch/issues/16267
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16270
Differential Revision:
D13791773
Pulled By: suo
fbshipit-source-id:
256304a02dbf724a7c0baade48c94b3ee77f53cf
Jesse Hellemn [Wed, 23 Jan 2019 23:35:00 +0000 (15:35 -0800)]
Fixing upload of nightly binaries and clean MacOS output (#16016)
Summary:
- Fix environment variable used to guard binary uploads
- Move common MacOS brew setup-code into a common function to decrease code duplication and also to move that noisy console output into its own CircleCI step
- Split Mac builds into separate build-test and upload jobs. Add one of these jobs to PR runs; add upload jobs to nightly binarybuilds workflow
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16016
Differential Revision:
D13791084
Pulled By: pjh5
fbshipit-source-id:
8eeb8e1963d46eab84f0f6dad9f0265163d5bf73
Teng Li [Wed, 23 Jan 2019 22:04:47 +0000 (14:04 -0800)]
CUDA event should only be recorded after NCCL group (#8219)
Summary:
Otherwise, it won't work if we sync on this event.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/8219
Reviewed By: pietern
Differential Revision:
D13788657
Pulled By: teng-li
fbshipit-source-id:
8c96e9691ed2441d7a685fb7ae8fece906f58daf
Edward Yang [Wed, 23 Jan 2019 21:51:05 +0000 (13:51 -0800)]
Change data() accessor in Caffe2 to return non-const pointer. (#16176)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16176
This makes PyTorch and Caffe2's data() method line up.
Historically, PyTorch made no distinction between tensors
with const or non-const data, and thus provided a
non-const pointer with data() member. Changing the API to
return a const-pointer would break all mutable code, whereas
changing the Caffe2 API to change a pointer doesn't break
any code, *except* for code which required an exact match
on const-ness (e.g., in template arguments). Since the latter
is less disruptive, we've opted for it here.
The few places downstream that broke due to this are fixed
in this patch.
Reviewed By: smessmer
Differential Revision:
D13742916
fbshipit-source-id:
baa4b4544cfdf7c1f369f4d69a1e0d5953c1bd99
svcscm [Wed, 23 Jan 2019 21:05:30 +0000 (13:05 -0800)]
Updating submodules
Reviewed By: cdelahousse
fbshipit-source-id:
99d58034f9369846f8c82a5ea11c71e202e52a4e
Christian Puhrsch [Wed, 23 Jan 2019 20:30:47 +0000 (12:30 -0800)]
Align native_functions.yaml func schema more with JIT signature schema (#16111)
Summary:
This PR applies a few minor modifications leading to 100s of additional matches
Modifications to native_functions.yaml
1) double to float
2) int64_t to int
3) IntList[\d*] to int[\d*]
4) {} to []
5) Tensor? x=[] to Tensor? x=None
6) TensorList to Tensor[]
7) 1e-x to 1e-0x
8) Generator* x = nullptr to Generator? x = None
9) `{.*}` to `[.*]`
Overall this adds about 300 new matches and brings us to about 1/2 compliance of native_functions func with their JIT signature equivalent
While this is still a draft "tools/jit/gen_jit_dispatch.py" contains code to aid in finding close signatures
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16111
Reviewed By: ezyang
Differential Revision:
D13738123
Pulled By: cpuhrsch
fbshipit-source-id:
d1ec1e089bdb26ec155f6f31ccf768270acb76c7
Syed Tousif Ahmed [Wed, 23 Jan 2019 20:29:32 +0000 (12:29 -0800)]
Fixes selection of cuDNN algorithm (#15881)
Summary:
This PR updates the logic for using cudnnGet* and cudnnFind*. Current version of cudnn find and get (v7) returns a pair of best algorithm and the convDesc mathType. While we were using the returned algorithm, we didn't update the mathType. As a result, we ended up with a slow choice of algorithm and math type. Without this patch, we are seeing a 10x regression in group convolutions.
Changelist:
- Changed the template arguments to be `perf_t` instead of `algo_t` to unify cudnnFind and cudnnGet. Both cudnnFind and cudnnGet have the same purpose and hence, it made sense to unify them and get rid of `getAlgorithm`.
- Used cudnnGet*_v7 everywhere cudnnGet* was being used.
- Removed all cudnn6 paths (This PR depends on https://github.com/pytorch/pytorch/pull/15851)
Differential Revision:
D13787601
Pulled By: ezyang
fbshipit-source-id:
81fe86727673d021306fe1c99c3e528b7c9ad17f
Edward Yang [Wed, 23 Jan 2019 19:53:55 +0000 (11:53 -0800)]
Disable flaky test
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16274
Reviewed By: pietern
Differential Revision:
D13788036
fbshipit-source-id:
a9b7353fb0655908e6d47387cc77af33e9471aed
Junjie Bai [Wed, 23 Jan 2019 17:31:14 +0000 (09:31 -0800)]
Update third_party protobuf to v3.6.1
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16251
Reviewed By: ezyang
Differential Revision:
D13781444
Pulled By: bddppq
fbshipit-source-id:
b713a021033d214f30a49ee02b95edf8633bcc50
Armaan Sethi [Wed, 23 Jan 2019 16:32:29 +0000 (08:32 -0800)]
fix sigma in the middle of when word (#16227)
Summary:
there is a random sigma in the when word on :
https://pytorch.org/cppdocs/contributing.html
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16227
Differential Revision:
D13762753
Pulled By: goldsborough
fbshipit-source-id:
3d4bf4be859a3069402fe8c3fbc8ebee4f25cc5a
Derek Kim [Wed, 23 Jan 2019 10:57:56 +0000 (02:57 -0800)]
Typos and broken RSTs fixed in torch.distribution (#16136)
Summary:
- probabilty -> probability
- make long lines break
- Add LogitRelaxedBernoulli in distribution.rst
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16136
Differential Revision:
D13780406
Pulled By: soumith
fbshipit-source-id:
54beb975eb18c7d67779a9631dacf7d1461a6b32
Johannes M Dieterich [Wed, 23 Jan 2019 02:21:07 +0000 (18:21 -0800)]
tune elementwise for AMD uarch (#16217)
Summary:
Tune elementwise kernel for AMD architectures by increasing the work group sizes and launch bounds. This change improves training throughput for torchvision models by up to 11% in our tests while exhibiting no significant performance regression.
No functional/performance change for CUDA - just shifting numbers into constrexpr.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16217
Differential Revision:
D13776684
Pulled By: bddppq
fbshipit-source-id:
edbaebe904598b2de66a9e9a68a1aa219ebc01e9
rohithkrn [Wed, 23 Jan 2019 01:19:15 +0000 (17:19 -0800)]
fix typo in resnet50_trainer.py
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16219
Differential Revision:
D13776742
Pulled By: bddppq
fbshipit-source-id:
10a6ab4c58159b3f619b739074f773662722c1d9
Lu Fang [Tue, 22 Jan 2019 22:55:50 +0000 (14:55 -0800)]
update of fbcode/onnx to
dc75285d4a1cff9618400164dfdb26c5a1bab70a
Summary:
Previous import was
c553fb32a0902ce5dd42e1b40123e9e9b38bdbe7
Included changes:
- **[dc75285](https://github.com/onnx/onnx/commit/dc75285)**: Relax constraint that the initializers must be a subset of graph inputs (#1718) <G. Ramalingam>
- **[985c8cd](https://github.com/onnx/onnx/commit/985c8cd)**: Fix typo in scan shape inferencing (#1753) <Scott McKay>
- **[ab52a5d](https://github.com/onnx/onnx/commit/ab52a5d)**: remove stale test cases <Lu Fang>
- **[56434bb](https://github.com/onnx/onnx/commit/56434bb)**: Removing experimental ConstantFill op. <Spandan Tiwari>
- **[881c63c](https://github.com/onnx/onnx/commit/881c63c)**: Show string names of data types instead of int IDs (#1749) <Shinichiro Hamaji>
- **[0a12fe4](https://github.com/onnx/onnx/commit/0a12fe4)**: Update ConstantOfShape op. (#1744) <Bowen Bao>
- **[ef028e5](https://github.com/onnx/onnx/commit/ef028e5)**: Update definition of Cast Op to support casting to/from string (#1704) <Raymond Yang>
Reviewed By: BIT-silence
Differential Revision:
D13773962
fbshipit-source-id:
b98079277994a699d4807210ba1d9c27f4672090
Shen Li [Tue, 22 Jan 2019 22:28:18 +0000 (14:28 -0800)]
Add default_stream() and enhance current_stream() (#16200)
Summary:
Closes #16156
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16200
Differential Revision:
D13747455
Pulled By: mrshenli
fbshipit-source-id:
00c0d5f341c3ac7a757bdb4631a17e11fbc6d3ec
Edward Yang [Tue, 22 Jan 2019 22:17:20 +0000 (14:17 -0800)]
complex_registration_extension.cpp includes to angled brackets
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16122
Reviewed By: smessmer
Differential Revision:
D13717900
fbshipit-source-id:
8401f39d993482d3e08d2d79bc1841deafee2a5b
Edward Yang [Tue, 22 Jan 2019 22:17:20 +0000 (14:17 -0800)]
Remove ATen/Allocator.h forwarding header.
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16121
Reviewed By: smessmer
Differential Revision:
D13717899
fbshipit-source-id:
83488f2aa801ca75059949ec85171ec03e64c4ff
Edward Yang [Tue, 22 Jan 2019 21:35:03 +0000 (13:35 -0800)]
Remove dead curVal store.
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16116
Reviewed By: smessmer
Differential Revision:
D13717719
fbshipit-source-id:
2ecee3f08f64e64ec5ac3c92fb326bc3df37e40e
Sebastian Messmer [Tue, 22 Jan 2019 21:21:38 +0000 (13:21 -0800)]
Make kernel registration constexpr again (#16166)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16166
Since we now don't use std::function anymore, we can make kernel registration constexpr again.
Reviewed By: ezyang
Differential Revision:
D13738630
fbshipit-source-id:
918fa3a3c8c6f0ddbd0f08b3b143cdf066265387
Sebastian Messmer [Tue, 22 Jan 2019 21:21:38 +0000 (13:21 -0800)]
Avoid closure around kernel (#16165)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16165
Store kernels as direct function pointers instead of std::function.
Using direct function pointers avoids a performance risk std::function would introduce.
Reviewed By: ezyang
Differential Revision:
D13738627
fbshipit-source-id:
a348906c8a201436699681980a82ca95065a06a0
Sebastian Messmer [Tue, 22 Jan 2019 21:21:37 +0000 (13:21 -0800)]
Pass IValues from JIT to c10 dispatcher (#16066)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16066
Don't unwrap and re-wrap but directly pass through the IValues
Reviewed By: ezyang
Differential Revision:
D13689037
fbshipit-source-id:
99b8155e640eb61a3c0597bf0f2b9c338712b45e
Shen Li [Tue, 22 Jan 2019 21:14:24 +0000 (13:14 -0800)]
Release GIL when synchronize or wait (#16182)
Summary:
address the second future work item in #15937
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16182
Differential Revision:
D13744972
Pulled By: mrshenli
fbshipit-source-id:
e9812e3fd4a5623e99b639d9f334bfc2d1827d92
Wanchao Liang [Tue, 22 Jan 2019 20:11:23 +0000 (12:11 -0800)]
Revert
D13540278: [pytorch][PR] Unhide unique from C++, make unique partially scriptable
Differential Revision:
D13540278
Original commit changeset:
3768c76a90b0
fbshipit-source-id:
7a31c239f9dca6ff467344d99820095addcae9d7
Xiang Gao [Tue, 22 Jan 2019 19:09:18 +0000 (11:09 -0800)]
Return namedtuples from torch.* function with multiple return arguments for C++ operators (#15429)
Summary:
Partially fixes: https://github.com/pytorch/pytorch/issues/394
Implementation detail:
Codegen is modified to generate codes that looks like below:
```C++
static PyObject * THPVariable_svd(PyObject* self_, PyObject* args, PyObject* kwargs)
{
HANDLE_TH_ERRORS
static PythonArgParser parser({
"svd(Tensor input, bool some=True, bool compute_uv=True, *, TensorList[3] out=None)",
}, /*traceable=*/true);
ParsedArgs<6> parsed_args;
auto r = parser.parse(args, kwargs, parsed_args);
static PyStructSequence_Field fields0[] = {
{"U", ""}, {"S", ""}, {"V", ""}, {nullptr}
};
static PyStructSequence_Desc desc0 = {
"torch.return_types.svd_out", nullptr,
fields0, 3
};
static PyTypeObject type0;
static bool namedtuple_type_initialized0 = false;
if (!namedtuple_type_initialized0) {
PyStructSequence_InitType(&type0, &desc0);
namedtuple_type_initialized0 = true;
}
static PyStructSequence_Field fields1[] = {
{"U", ""}, {"S", ""}, {"V", ""}, {nullptr}
};
static PyStructSequence_Desc desc1 = {
"torch.return_types.svd", nullptr,
fields1, 3
};
static PyTypeObject type1;
static bool namedtuple_type_initialized1 = false;
if (!namedtuple_type_initialized1) {
PyStructSequence_InitType(&type1, &desc1);
namedtuple_type_initialized1 = true;
}
if (r.idx == 0) {
if (r.isNone(3)) {
return wrap(&type1, dispatch_svd(r.tensor(0), r.toBool(1), r.toBool(2)));
} else {
auto results = r.tensorlist_n<3>(3);
return wrap(&type0, dispatch_svd(r.tensor(0), r.toBool(1), r.toBool(2), results[0], results[1], results[2]));
}
}
Py_RETURN_NONE;
END_HANDLE_TH_ERRORS
}
```
Types are defined as static member of `THPVariable_${op_name}` functions, and initialized at the first time the function is called.
When parsing function prototypes in `native_functions.yaml`, the parser will set the specified name as `field_name` when see things like `-> (Tensor t1, ...)`. These field names will be the field names of namedtuple. The class of namedtuples will be named `torch.return_types.${op_name}`.
In some python 2, `PyStructSequence` is not a subtype of tuple, so we have to create some functions to check if an object is a tuple or namedtuple for compatibility issue.
Operators in `native_functions.yaml` are changed such that only `max` and `svd` are generated as namedtuple. Tests are added for these two operators to see if the return value works as expected. Docs for these two ops are also updated to explicitly mention the return value is a namedtuple. More ops will be added in later PRs.
There is some issue with Windows build of linker unable to resolve `PyStructSequence_UnnamedField`, and some workaround is added to deal with this case.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15429
Differential Revision:
D13709678
Pulled By: ezyang
fbshipit-source-id:
23a511c9436977098afc49374e9a748b6e30bccf
Jongsoo Park [Tue, 22 Jan 2019 18:08:33 +0000 (10:08 -0800)]
Fix formating in caffe2/quantization/server/README.md
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/14237
Reviewed By: dskhudia
Differential Revision:
D13751791
Pulled By: jspark1105
fbshipit-source-id:
54f73d5134e596817802c66d43098d18458c2799
Yaxun (Sam) Liu [Tue, 22 Jan 2019 17:07:18 +0000 (09:07 -0800)]
hip-clang enablement (#16085)
Summary:
Initial enabling of the upcoming hip-clang compiler for the PyTorch source base.
Changes:
* update the Eigen submodule to a version including our upstreamed hip-clang enabling there
* modify a few ifdef guards with the `__HIP__` macro used by hip-clang
* use `__lane_id` instead of `hc::__lane_id`
* add Debug flags for ROCm to the cmake infrastructure
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16085
Differential Revision:
D13709459
Pulled By: ezyang
fbshipit-source-id:
1b7b33fe810a0434766180580d4443ea177eb7c7
Andy Wei [Tue, 22 Jan 2019 16:46:06 +0000 (08:46 -0800)]
Raise CalledProcessError when torch.distributed launch process not return 0 (#16069)
Summary:
`torch.distributed.launch.py` will not raise error when `subprocess.Popen` is not return 0.
For better debugging it should always raise an error if processes launched have unusual behavior
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16069
Differential Revision:
D13709467
Pulled By: ezyang
fbshipit-source-id:
31d32a5ec8fed7bccd62d845bfba0e670ed3fe20
Shahzad Lone [Tue, 22 Jan 2019 16:00:00 +0000 (08:00 -0800)]
Reserve vectors that we know the size in advance for. (#16201)
Summary:
Save reallocation costs, by reserving vectors according to how many elements we expect to put in.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16201
Differential Revision:
D13762594
Pulled By: ezyang
fbshipit-source-id:
7e3bfe421489dde48a2ddb0920dd155f69baecc0
Will Feng [Tue, 22 Jan 2019 05:53:43 +0000 (21:53 -0800)]
cpp doc fix (#16221)
Summary:
Fixed a few C++ API callsites to work with v1.0.1.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16221
Differential Revision:
D13759207
Pulled By: yf225
fbshipit-source-id:
bd92c2b95a0c6ff3ba5d73cb249d0bc88cfdc340
Lu Fang [Tue, 22 Jan 2019 04:13:07 +0000 (20:13 -0800)]
Move away from ConstantFill (#16214)
Summary:
Prerequisite of https://github.com/onnx/onnx/pull/1434
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16214
Reviewed By: BIT-silence
Differential Revision:
D13755116
Pulled By: houseroad
fbshipit-source-id:
a46be8d7df959b5ede93e1f9c911a9a9326e6879
Zachary DeVito [Tue, 22 Jan 2019 03:57:33 +0000 (19:57 -0800)]
ban conv_double_backward from sandcastle, it takes too long
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/16220
Differential Revision:
D13755108
Pulled By: zdevito
fbshipit-source-id:
46b1b128b155964c25249add0c84680491845e9b
Zachary DeVito [Tue, 22 Jan 2019 01:24:32 +0000 (17:24 -0800)]
Remove dead code from setup.py, remove need for build target. (#16162)
Summary:
Now it is only necessary to use 'develop' or 'install' to build. Incremental cmake is on by default. `develop --cmake` forces it to rerun.
The NinjaBuilder stuff is dead. It was used to make building _C.so
faster but now _C.so is just an empty stub file.
Removed a bunch of custom build commands from setup.py that are
no longer meaningful now that cmake handles most of the build.
Removed unused targets in build_pytorch_lib.sh/bat
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16162
Differential Revision:
D13744155
Pulled By: zdevito
fbshipit-source-id:
d836484782c65b7f8e8c7a82620886f7a7777892
Xiang Gao [Mon, 21 Jan 2019 20:28:56 +0000 (12:28 -0800)]
Unhide unique from C++, make unique partially scriptable (#15256)
Summary:
This PR does three things:
~~Allow `int64_t?` in function schema, which provide an elegant way of implementing null-able int arguments, as discussed in https://github.com/pytorch/pytorch/pull/15208#pullrequestreview-
185230081~~
~~Originally implemented in https://github.com/pytorch/pytorch/pull/15235~~
~~Example:~~
```yaml
- func: myop(Tensor self, int64_t? dim=None) -> Tensor
variants: function
```
~~cc: zou3519~~
Edit: implemented in https://github.com/pytorch/pytorch/pull/15234
Previously tried in https://github.com/pytorch/pytorch/pull/12064. There was a problem that C++ does not have kwarg support, which makes it confusing to know whether `unique(t, 1)` actually means `unique(t, dim=1)` or `unique(t, sorted=1)`.
Now I think I have a better idea on how to implement this: there are two ATen operators: `unique` and `unique_dim`. `unique` has the same signature as in python, and exported to both python and C++. `unique_dim` has signature `unique_dim(tensor, dim, sorted=False, return_inverse=False)`, and only exported to C++, which could be used more naturally for a C++ user.
Differential Revision:
D13540278
Pulled By: wanchaol
fbshipit-source-id:
3768c76a90b0881f565a1f890459ebccbdfe6ecd
Lu Fang [Mon, 21 Jan 2019 17:43:40 +0000 (09:43 -0800)]
update of fbcode/onnx to
c553fb32a0902ce5dd42e1b40123e9e9b38bdbe7 (#16190)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16190
Previous import was
fd60104394fa353e1762f44ecad1b2166e33deef
Included changes:
- **[c553fb3](https://github.com/onnx/onnx/commit/c553fb3)**: Handle negative axis in scan shape inference (#1748) <G. Ramalingam>
- **[51b6ecc](https://github.com/onnx/onnx/commit/51b6ecc)**: external_data: Store large tensor values in separate files (#678) <Michał Karzyński>
- **[ba05f26](https://github.com/onnx/onnx/commit/ba05f26)**: Scan output axes (#1737) <G. Ramalingam>
- **[90920c0](https://github.com/onnx/onnx/commit/90920c0)**: Add NonZero op. (#1714) <Sergii Dymchenko>
- **[c4cf112](https://github.com/onnx/onnx/commit/c4cf112)**: fix the test cases for constantofshape (#1746) <Lu Fang>
- **[d902349](https://github.com/onnx/onnx/commit/d902349)**: Add sample implementation support (#1712) <Lu Fang>
Differential Revision:
D13745693
fbshipit-source-id:
05e2cce9ae1dfa2865db83840df64673d55cea57