Summary:
The old `torch.distributed` will go to `torch.distributed.deprecated`
The old DDP will go to `torch.nn.parallel.deprecated`
Now `torch.nn.parallel.DDP` will use c10d DDP
Now `torch.distributed` will use C10d frontend API
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11405
Reviewed By: pietern
Differential Revision: D9733733
Pulled By: teng-li
fbshipit-source-id: d6a3f3e73f8d3a7fcb1f4baef53c78063b8cbb08
Summary:
Example:
```sh
python run_test.py -i sparse -- TestSparse.test_factory_size_check -f
```
With this, the `--verbose` option is redundant (one can call `python run_test.py -- -v` instead of `python run_test.py -v`. But since this is (probably) a frequently used flag, I didn't remove the existing easier-to-use option.
cc ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11209
Differential Revision: D9632215
Pulled By: SsnL
fbshipit-source-id: ff522802da11ef0a0714578be46e4a44f6343d44
Summary:
* improve docker packages (install OpenBLAS to have at-compile-time LAPACK functionality w/ optimizations for both Intel and AMD CPUs)
* integrate rocFFT (i.e., enable Fourier functionality)
* fix bugs in ROCm caused by wrong warp size
* enable more test sets, skip the tests that don't work on ROCm yet
* don't disable asserts any longer in hipification
* small improvements
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10893
Differential Revision: D9615053
Pulled By: ezyang
fbshipit-source-id: 864b4d27bf089421f7dfd8065e5017f9ea2f7b3b
Summary:
The PR includes:
(1) torch.distributed.c10d, which now includes the complete backward compatible frontend API for `torch.distributed`
(2) `env://` init method functionality
(3) Minor change to `test_distributed.py`, which is now a test for `torch.distributed.c10d`.
(4) The old `test_distributed.py' is now moved to `test_distributed_thd`
(5) Miscellaneous bug fixes.
(6) DDP CPU test is removed since c10d doesn't have this support yet, but this is a very easy test after moving DDP CPU's dependency to torch.distributed.c10d.
(7) CI config to test MPI, NCCL, and Gloo backend of c10d
**Now all the distributed test including c10d DDP can pass with the c10d frontend API**
TODO: (in a separate PR)
MPI subgroup support, once this is added, CI group test will be enabled.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10871
Differential Revision: D9554514
Pulled By: teng-li
fbshipit-source-id: fb686ad42258526c8b4372148e82969fac4f42dd
Summary:
The previous NCCL all gather doesn't work as expected. This is a fully working async version. Tested on both C++ and Python Frontend.
Multi-node:
```
tengli@learnfair042:~/new_pytorch/pytorch/torch/lib/build/c10d/test$ TMPFILE="/private/home/tengli/temp/tengli-test" RANK=0 WORLD_SIZE=2 ./ProcessGroupNCCLTest
Multi-node world size: 2 rank: 0
Allreduce test successful
Broadcast test successful
Reduce test successful
Allgather test successful
tengli@learnfair117:~/new_pytorch/pytorch/torch/lib/build/c10d/test$ TMPFILE="/private/home/tengli/temp/tengli-test" RANK=1 WORLD_SIZE=2 ./ProcessGroupNCCLTest
Multi-node world size: 2 rank: 1
Allreduce test successful
Broadcast test successful
Reduce test successful
Allgather test successful
```
CI test:
```
test_set_get (__main__.FileStoreTest) ... ok
test_set_get (__main__.PrefixFileStoreTest) ... ok
test_set_get (__main__.PrefixTCPStoreTest) ... ok
test_allreduce_ops (__main__.ProcessGroupGlooTest) ... ok
test_broadcast_ops (__main__.ProcessGroupGlooTest) ... ok
test_allgather_ops (__main__.ProcessGroupNCCLTest) ... ok
test_allreduce_ops (__main__.ProcessGroupNCCLTest) ... ok
test_broadcast_ops (__main__.ProcessGroupNCCLTest) ... ok
test_reduce_ops (__main__.ProcessGroupNCCLTest) ... ok
test_common_errors (__main__.RendezvousFileTest) ... ok
test_nominal (__main__.RendezvousFileTest) ... ok
test_common_errors (__main__.RendezvousTCPTest) ... ok
test_nominal (__main__.RendezvousTCPTest) ... ok
test_unknown_handler (__main__.RendezvousTest) ... ok
test_set_get (__main__.TCPStoreTest) ... ok
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10932
Differential Revision: D9542067
Pulled By: teng-li
fbshipit-source-id: 25513eddcc3119fd736875d69dfb631b10f4ac86
Summary:
* first integration of MIOpen for batch norm and conv on ROCm
* workaround a ROCm compiler bug exposed by elementwise_kernel through explicit capture of variables in the densest packing
* workaround a ROCm compiler bug exposed by having `extern "C" __host__` as a definition and just `__host__` in the implementation through the hipify script
* use fabs() in accordance with C++11 for double absolute, not ::abs() which is integer-only on ROCm
* enable test_sparse set on CI, skip tests that don't work currently on ROCm
* enable more tests in test_optim after the elementwise_bug got fixed
* enable more tests in test_dataloader
* improvements to hipification and ROCm build
With this, resnet18 on CIFAR data trains without hang or crash in our tests.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10612
Reviewed By: bddppq
Differential Revision: D9423872
Pulled By: ezyang
fbshipit-source-id: 22c0c985217d65c593f35762b3eb16969ad96bdd
Summary:
* some small leftovers from the last PR review
* enable more unit test sets for CI
* replace use of hcRNG w/ rocRAND (docker image was already updated w/ newer rocRAND)
* use rocBLAS instead of hipBLAS to allow convergence w/ Caffe2
* use strided_batched gemm interface also from the batched internal interface
* re-enable Dropout.cu as we now have philox w/ rocRAND
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10406
Reviewed By: Jorghi12
Differential Revision: D9277093
Pulled By: ezyang
fbshipit-source-id: 7ef2f6fe4ead77e501ed7aea5c3743afe2466ca2
Summary:
This PR for the ROCm target does the following:
* enable some unit tests on ROCm
* fix a missing static_cast that breaks BatchNorm call on ROCm
* fix BatchNorm to work on ROCm w/ ROCm warp sizes etc
* improve the pyhipify script by introducing kernel scope to some transpilations and other improvements
* fix a linking issue on ROCm
* for more unit test sets: mark currently broken tests broken (to be fixed)
* enable THINLTO (phase one) to parallelize linking
* address the first failing of the elementwise kernel by removing non-working ROCm specialization
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10266
Differential Revision: D9184178
Pulled By: ezyang
fbshipit-source-id: 03bcd1fe4ca4dd3241f09634dbd42b6a4c350297
* Build and install c10d from tools/build_pytorch_libs.sh
* Create initial Python bindings for c10d
* clang-format
* Switch link order to include more symbols
* Add bindings and tests for ProcessGroupGloo
* Add broadcast test
* Separate build flag for c10d
* Explicit PIC property
* Skip c10d tests if not available
* Remove c10d from Windows blacklist
Let it skip by itself because it won't be available anyway.
* Make lint happy
* Comments
* Move c10d module into torch.distributed
* Close tempfile such that it is deleted
* Add memory leak check in CUDA tests
* Tracking multi-GPU too
* fix run_test.py not running __name__ == '__main__' content; add test for make_cuda_memory_checked_test
* add a comment
* skip if cuda
* 1. Change the wrapper to a method in common.py:TestCase
2. Refactor common constants/method that initialize CUDA context into common_cuda.py
3. Update some test files to use TEST_CUDA and TEST_MULTIGPU
* Fix MaxUnpool3d forward memory leak
* Fix MultiLabelMarginCriterion forward memory leak
* Fix MultiMarginLoss backward memory leak
* default doCUDAMemoryCheck to False
* make the wrapper skip-able
* use TEST_MULTIGPU
* add align_corners=True/False tests for Upsample; fix TEST_CUDNN
* finalize interface
* VolumetricMaxUnpooling_updateOutput
* fix test_nccl
* rename THC caching allocator methods to be clearer
* make the wrapped function a method
* address comments; revert changes to aten/src/THC/THCCachingAllocator.cpp
* fix renamed var
* Add support for dotted names in CPP Extensions
* Modify tests for cpp extensions
Test that dotted names work
* Py2 fixes
* Make run_test cpp_extensions Win-compatible
* Check for --noprefix option for mpiexec
--noprefix option to mpiexec is not part of the MPI standard.
It is needed in certain configurations when using OpenMPI but not
supported with other MPI implementations such as MPICH and maybe
others. This commit adds a check if the option is supported by
the current mpiexec. Also this commit fixes Issue #4965 and MPI
tests can be enabled in the CI.
Fixes: #4965
* Update run_test.py
* Improve run_test.py to support running individual test classes and methods
Added support in run_test.py for running individual test classes and methods.
The -i/--include option can specify a list of test modules, classes or methods
like this:
python run_test.py -i autograd torch.TestTorch.test_abs \
torch.TestTorch.test_add utils.TestBottleneck
-f, -l and -x behaviour stays the same as before
* Fixed some code formatting
* Multiple fixes according to the reviews in #6344
* Change cpp_extensions.py to make it work on Windows
* Fix linting
* Show python paths
* Debug
* Debug 1
* set PYTHONPATH
* Add ATen into library
* expose essential libs and functions, and copy _C.lib
* Specify dir in header
* Update check_abi for MSVC
* Activate cl environment to compile cpp extensions
* change version string
* Redirect stderr to stdout
* Add monkey patch for windows
* Remove unnecessary self
* Fix various issues
* Append necessary flags
* add /MD flag to cuda
* Install ninja
* Use THP_API instead of THP_CLASS
* Beautify the paths
* Revert "Use THP_API instead of THP_CLASS"
This reverts commit dd7e74c44db48e4c5f85bb8e3c698ff9de71ba2d.
* Use THP_API instead of THP_CLASS(new)
- All of the scripts are based off of the idea that they should be as
simple as possible, and all the heavy lifting done in the construction
of the Docker file. The scripts are really simple now. A bigger
philosophical discussion can be found in .jenkins/README.md
- build-asan.sh is split out of build.sh, as ASAN builds are a bit
specialized and it's inappropriate to run many of the other builds
as part of them.
- We now build and run with mkl/mkl-include on the CPU only builds
- We now report sccache and ccache stats at the end of all builds.
- run_test.py flushes stdout/stderr before making a subprocess call,
which should solve our interleaving problems.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
* Simplify run_test.py and dont use shell=True
* Fix non-shell output for check_output and always print to stderr
* Use shlex.split instead of str.split
* s/log/print_to_stderr
* with_init -> with_init_file
* Remove bufsize argument
I need this because run_test is going to need to read other
options than just verbose when I implement JUnit XML dumping.
(JUnit XML dumping cannot be implemented solely by frobbing
--python because the XML file to dump to must vary based on the
test name.)
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
* Revert "ATen ReduceOps (#5481)"
This reverts commit 310c3735b9.
* Revert "Check that new cpuinfo and tbb submodules exist (#5714)"
This reverts commit 1a23c9901d.
This diff adds vectorization to ATen. It uses intel intrinsics to build a general vec256 class, that represents types of 256bit width. These can then be treated like regular variables. Using those it implements torch.sum() for the contiguous case. It uses Intel TBB for multithreading, which allows workstealing and chunks the reduction operations based on a experimentally chosen value (_THRESHOLD). It uses cpuinfo to pick the right code depending on the host's capabilities.
The kernels are implemented under native/cpu. Each .cpp file is compiled with -avx, -avx2 and no additional flags. A macro is used to append AVX, AVX2 or NONE to the function name. The header then needs to define the functions three times, one for each capability. This could be improved by either changing the cmake file a bit or possibly generating source code using a Python script etc.
For the non-contiguous case this defaults to the current implementation within TH. For CUDA is entirely defaults to the implementation within THC.
There probably needs to be a bit of a debate around the design decisions here, the additional dependencies, parallelization strategy, clarity, etc. The numerical results also diverge from numpy with larger tensors, which is expected since we're summing, for example, 8 numbers and then adding the result to the running sum, instead of each number one by one. But there might be something to be said about accumulating into a double for floats or the degree of divergence, the behavior with respect to CUDA, etc.
I wrote a [small Python script]( https://github.com/cpuhrsch/benchmark/blob/sumall/benchmarks/sum_bench.py) to compare the results with numpy numerically as well as on timing. I ran this script to create timings both on master and this branch.
Here is the command for 1 core
`OMP_NUM_THREAD=1 taskset -c 0 python sum_bench.py --enable_numpy 200`
Here is the command for all cores
`python sum_bench.py --enable_numpy 200`
Here are the results of each:
[Master, 1 core](https://paste.fedoraproject.org/paste/Nho9JzHpPVK9av8a6mByjQ)
[This branch, 1 core](https://paste.fedoraproject.org/paste/6xLHkYvcVJx9z~5MoHxN4w)
[Master, all cores](https://paste.fedoraproject.org/paste/5l3V1d5zGqvJcMXIUteMRw)
[This branch, all cores](https://paste.fedoraproject.org/paste/J4RuDU-0Drz0aZwtphQwEA)
To test the command is
`python sum_bench.py --test 200`
[This branch, test results](https://paste.fedoraproject.org/paste/kTEoUC~oWgXA6XWMAfNfNw)
For this test we look at the average absolute value of the differences. This does not take into account the relative magnitude of the numbers. The numbers are sampled from a standard normal distribution.
In terms of performance this diff should bring PyTorch on par with Numpy and usually exceed it by 1.5 to 2x.
