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
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
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16175
Separate Moments from math and optimize it
i-am-not-moving-c2-to-c10
Reviewed By: houseroad
Differential Revision: D13742472
fbshipit-source-id: 90757d908d38c98ca69818855aaf68315e525992
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16135
Separate affine_channel from math and optimize it
i-am-not-moving-c2-to-c10
Reviewed By: houseroad
Differential Revision: D13727606
fbshipit-source-id: 8980af4afadaf964a18a9da581106fe30896a7e9
Summary:
This tests the water for adding back NNPACK in PyTorch, it's a lot better than the fallback THNN versions.
In #6151, we (ezyang and soumith) removed NNPACK support from PyTorch. Of course Maratyszcza might have advice, too. (Or an opinion on the CMake changes.)
The only functional changes are to use NNPack more aggressively on mobile and a .contiguous() to match NNPack's assumption (I stumbled over that while using NNPack for style transfer.)
The CMake changes try to use the NNPack we already have in git.
In terms of lines of code this is a large part of the diff of https://lernapparat.de/pytorch-jit-android/ . As far as I can tell, we don't have MKLDNN on mobile and the native THNN implementation are prohibitively expensive in terms of both CPU and memory.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15924
Differential Revision: D13709576
Pulled By: ezyang
fbshipit-source-id: f2e287739909451c173abf046588209a7450ca2c
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15884
Codemod generated with clangr shard mode, 25 files per diff,
To eliminiate partially initialized Tensor, we split the initialization of local Tensor variables into two steps, first declare un uninitialized Tensor, and
call `ReinitializeTensor` to initialize it.
motivation: https://github.com/pytorch/pytorch/pull/12407
Reviewed By: hyuen
Differential Revision: D13586737
fbshipit-source-id: dc8e49e9f29505b8898bb19f84c1a983f2d811ab
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15316
This starts cleaning up the files in c10 according to the module structure we decided on.
Move to c10/util:
- Half.h, Half-inl.h, Half.cpp, bitcasts.h
Move to c10/core:
- Device.h, Device.cpp
- DeviceType.h, DeviceType.cpp
i-am-not-moving-c2-to-c10
Reviewed By: dzhulgakov
Differential Revision: D13498493
fbshipit-source-id: dfcf1c490474a12ab950c72ca686b8ad86428f63
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
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14918
When ProtoBuf-Lite is in use, ProtoDebugString just calls SerializeAsString.
This produces binary output, which is not a very suitable "debug" string.
Specifically, we've observed it causing problems when calling code tries to
add the debug string to a Java exception message (which requires valid UTF-8).
Now, we replace all non-ASCII bytes with "?".
This is not a very fast implementation, but generating debug strings shouldn't
be a performance-sensitive operation in any application.
Reviewed By: dzhulgakov
Differential Revision: D13385540
fbshipit-source-id: 8868172baf20efaf53fecf7d666a6980f59b64f5
Summary:
(1) Move Caffe2 thread pool to aten
(2) Use the same thread pool definition for PyTorch interpreter
(3) Make ivalue::Future thread-safe
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14114
Reviewed By: ilia-cher
Differential Revision: D13110451
Pulled By: highker
fbshipit-source-id: a83acb6a4bafb7f674e3fe3d58f7a74c68064fac
Summary:
Hi guys,
I'd like to build Caffe2 with more supported options in Windows with Microsoft Visual Studios.
This is the first pull request.
Running scripts/build_windows_shared.bat is able to build Caffe2 with both CMAKE_BUILD_TYPE=Debug and CMAKE_BUILD_TYPE=Release with Visual Studio 14 2015.
CUDA is 9.0, cudnn is 7.0.5, glog, gflags and lmdb are supported on my system.
Python is 3.5, Detectron works from python interface as well.
It was even possible to debug detectron code and step into caffe2_gpu.dll with pdbs built.
What is disappointing, that c10/experimental ops don't build with this Visual Studio generator, I added special option INCLUDE_EXPERIMENTAL_C10_OPS (default ON) to deal with it in build_windows_shared.bat.
After this pull request the next step is to add Visual Studio 2017 support in the script.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13550
Reviewed By: ezyang
Differential Revision: D13042597
Pulled By: orionr
fbshipit-source-id: f313f909f599cd582a1d000eff766eef3a9fc4fc
Summary:
xw285cornell
- To make hip files to have unique filename extension we change hip files from _hip.cc to .hip (it's the only blessing option other than .cu in hipcc 3d51a1fb01/bin/hipcc (L552)).
- Change to use host compiler to compile .cc|.cpp files. Previously we use hcc to compile them which is unnecessary
- Change the hipify script to not replace "gpu" with "hip" in the filename of the generated hipified files. Previously we do this because hcc has a bug when linking files that have same filename. We have now changed to use host linker to do linking so this is unnecessary anymore.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14036
Reviewed By: xw285cornell
Differential Revision: D13091813
Pulled By: bddppq
fbshipit-source-id: ea3d887751d8abb39d75f5d5104aa66ce66b9ee0
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13949
This diff adds support to fillers for `SparseLengthsWeight*` ops. It does 3 things:
1. Add the fillers for `SparseLengthsWeight*` ops
2. Add filling heuristics to consider the path of `LengthsRangeFill` -> `Gather` -> `SparseLengthsWeightedSum`, where the length input is shared by `LengthsRangeFill` and `SparseLengthsWeightedSum`. Therefore, we need to carefully bound the value of that length input so that at `Gather`, it does not index out-of-bound for the weight input of `Gather`.
3. Fix and simplify the logic of `math::RandFixedSum`, where we just keep rejecting the generated value if it violates the invariants.
Reviewed By: highker
Differential Revision: D13048216
fbshipit-source-id: bfe402e07e6421b28548047d18b298c148e0ec87
Summary:
This PR contains changes for:
1. Adding HIP top_k operator in Caffe2
2. Added HIP equivalent definitions of GPUDefs and GPUScanUtils
3. Removing the top_k operator test from ROCm test ignore list
4. Bug fixes in related code in THC/THCAsmUtils.cuh
Differential Revision: D12986451
Pulled By: bddppq
fbshipit-source-id: 6d5241fb674eaeb7cde42166426ac88043b83504
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12733
Conv in NHWC layout only works for 2D images. This has been a pain point when implementing quantized 3D convolution because we need NHWC layout for best performance (note that NHWC layout in general gives better performance in CPU not just for quantized operators). For example, our quantized ops have a functionality to measure quantized error operator by operator but this needs running a shadow fp32 operator, but this is not easy when there's no 3D conv in NHWC layout is available (currently we're doing layout conversion on the fly for the shadow fp32 operator which is error prone). Some of Caffe2 frameworks like brew generates error when we try to create a 3D conv op in NHWC layout. This was also a blocker for using aibench because aibench is using brew.
i-am-not-moving-c2-to-c10
Reviewed By: houseroad
Differential Revision: D10333829
fbshipit-source-id: 2d203ee1db833cd3f9d39353219e3894b46c4389
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13480
Revert D12845220 since the MKL functions are using multi-thread while the single-thread run is slower than eigen version.
i-am-not-moving-c2-to-c10
Reviewed By: houseroad
Differential Revision: D12891751
fbshipit-source-id: 2a61727b269a304daeee2af6ff7fee7820cb5344
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11884
In cpu mode, current convNd uses Im2ColNdNCHWImpl, which is generic implementation to handle convolutional layer for arbitrary number of dimensions. In video modeling, we use convNd for filter dimension=3.
The problem of current convNd is that Im2ColNdNCHWImpl is much slower than Im2Col used by conv2d for the filters with same Flops. For example, a (1, 7, 7) 3d filter takes 5 times longer than a (7, 7) 2d filter at inference time.
This diff extends Im2Col to 3d case (Im2Col3dNCHWImpl), and this optimization for 3d convolution gives 4~5 times faster inference time on cpu for various video models:
{F128300920}
i-am-not-moving-c2-to-c10
Reviewed By: BIT-silence
Differential Revision: D8245940
fbshipit-source-id: 75231d65c9dd56059dfe31701e26021fd1ff2a85
Summary:
We lost HIP support in last refactoring 620ece2668
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13400
Differential Revision: D12868211
Pulled By: bddppq
fbshipit-source-id: 72dbfda105b826bee28ddf480e88fca7d63f93d8
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13205
CopyFrom without context argument does the sync copy on the current gpu - exactly what most of the places need.
This diff kills about 60% of CopyFrom usages. Most common pattern is gpu->cpu copy with further FinishDeviceComputation - the latter can be just killed.
Reviewed By: Yangqing
Differential Revision: D11236076
fbshipit-source-id: eb790ca494dfc5d5e3a7d850b45d6f73221bb204
Summary:
The experimental ops for the c10 dispatcher have accidentally been disabled in the oss build when the directory changed from `c10` to `experimental/c10`. This PR re-enables them.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12821
Differential Revision: D10446779
Pulled By: smessmer
fbshipit-source-id: ac58cd1ba1281370e62169ec26052d0962225375
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13114
Using one thread pool creator for all device types
Reviewed By: manojkris, wesolwsk
Differential Revision: D10851533
fbshipit-source-id: 32ca51d7932ba7faa8137df26315f52ecb4c6157
Summary:
Now we have everything from c10::optional, we can delete this and keep a single version in c10.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12965
Differential Revision: D10504042
Pulled By: wanchaol
fbshipit-source-id: c0ec3892e92968cca264ae8924c19111674631ba
Summary:
The legal function cublasHandle_t cublas_handle() was hipified to the
clearly illegal rocblas_handle rocblas_handle(). It should not work and
correctly fails with gcc as the host compiler as it induces an
ambiguity.
Function now hipifies to rocblas_handle rocblashandle()
Fixes long standing issue we've observed in PyTorch when base compiler is gcc.
For attention: bddppq ezyang
Tests on ROCm PyTorch/Caffe2: https://github.com/ROCmSoftwarePlatform/pytorch/pull/284
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12957
Differential Revision: D10501227
Pulled By: bddppq
fbshipit-source-id: 568cb80801c0d14c9b1b61e3a7db387a5c21acf4
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12714
This is a short change to enable c10 namespace in caffe2. We did not enable
it before due to gflags global variable confusion, but it should have been
mostly cleaned now. Right now, the plan on record is that namespace caffe2 and
namespace aten will fully be supersets of namespace c10.
Most of the diff is codemod, and only two places of non-codemod is in caffe2/core/common.h, where
```
using namespace c10;
```
is added, and in Flags.h, where instead of creating aliasing variables in c10 namespace, we directly put it in the global namespace to match gflags (and same behavior if gflags is not being built with).
Reviewed By: dzhulgakov
Differential Revision: D10390486
fbshipit-source-id: 5e2df730e28e29a052f513bddc558d9f78a23b9b
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12306
In a future diff, I'm going to introduce non-placement constructor and destructor to TypeMeta.
To make it less ambigous, this diff is first renaming the existing ones to PlacementXXX.
Reviewed By: dzhulgakov
Differential Revision: D10184117
fbshipit-source-id: 119120ebc718048bdc1d66e0cc4d6a7840e666a4
Summary:
There are still a few work to be done:
- Move logging and unify AT_WARN with LOG(ERROR).
- A few header files are still being plumbed through, need cleaning.
- caffe2::EnforceNotMet aliasing is not done yet.
- need to unify the macros. See c10/util/Exception.h
This is mainly a codemod and not causing functional changes. If you find your job failing and trace back to this diff, usually it can be fixed by the following approaches:
(1) add //caffe2/c10:c10 to your dependency (or transitive dependency).
(2) change objects such as at::Error, at::Optional to the c10 namespace.
(3) change functions to the c10 namespace. Especially, caffe2::MakeString is not overridden by the unified c10::str function. Nothing else changes.
Please kindly consider not reverting this diff - it involves multiple rounds of rebasing and the fix is usually simple. Contact jiayq@ or AI Platform Dev for details.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12354
Reviewed By: orionr
Differential Revision: D10238910
Pulled By: Yangqing
fbshipit-source-id: 7794d5bf2797ab0ca6ebaccaa2f7ebbd50ff8f32
