Libtorch is slower on Windows. Also, can't compile with MKLDNN

C++
1

Hi,

This is my first time posting so apologize if duplicate.

I am having an issue where my model is running much slower in windows than in linux. I believe this has to do with inclusion of MKLDNN when building. When I try to build in windows I am getting the following.

– Building version 1.4.0a0+7f73f1d
cmake -GNinja -DBUILD_PYTHON=True -DBUILD_TEST=True -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=C:\pytorch\torch -DCMAKE_PREFIX_PATH=C:\pytorch1.4\Lib\site-packages -DNUMPY_INCLUDE_DIR=C:\pytorch1.4\lib\site-packages\numpy\core\include -DPYTHON_EXECUTABLE=C:\pytorch1.4\python.exe -DPYTHON_INCLUDE_DIR=C:\pytorch1.4\include -DPYTHON_LIBRARY=C:\pytorch1.4/libs/python37.lib -DTORCH_BUILD_VERSION=1.4.0a0+7f73f1d -DUSE_NUMPY=True C:\pytorch
cmake --build . --target install --config Release – -j 48
Traceback (most recent call last):
File “setup.py”, line 755, in
build_deps()
File “setup.py”, line 316, in build_deps
cmake=cmake)
File “C:\pytorch\tools\build_pytorch_libs.py”, line 62, in build_caffe2
cmake.build(my_env)
File “C:\pytorch\tools\setup_helpers\cmake.py”, line 335, in build
self.run(build_args, my_env)
File “C:\pytorch\tools\setup_helpers\cmake.py”, line 141, in run
check_call(command, cwd=self.build_dir, env=env)
File “C:\pytorch1.4\lib\subprocess.py”, line 363, in check_call
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command ‘[‘cmake’, ‘–build’, ‘.’, ‘–target’, ‘install’, ‘–config’, ‘Release’, ‘–’, ‘-j’, ‘48’]’ returned non-zero exit status 1.
(env_pytorch1.4)

Any suggestions on how to resolve this? Thanks for your help in advance.

-Nick

2

The error message is rather useless. Would you please post the full log here? BTW, we added MKLDNN support for Windows in 1.5.0 so you may need to select master as the target branch. What’s more, you may try out this one.

3

I’ve tried building the latest and it seems to be completing further. It doesn’t compile every time it seems but generally it can make it to end of caffe2. I am getting the following error when I try to import torch.

importing torch
importing torch1118×334 12.3 KB

The following is the kdiff between torch/lib in a working (python installed version) vs master installed via python setup.py install.

kdiff_diffs_pip_vs_built
kdiff_diffs_pip_vs_built437×581 7.78 KB

would appreciate any advice.

Thanks,
Nick

4

Also, is there a reason why the 1.4.0 and 1.4.0a0 would be failing. I am attaching a porting of the logs from earlier (not sure how to attach the entire log).

Thanks,
Nick

– pytorch is compiling with OpenMP.
OpenMP CXX_FLAGS: -openmp:experimental.
OpenMP libraries: .
– Caffe2 is compiling with OpenMP.
OpenMP CXX_FLAGS: -openmp:experimental.
OpenMP libraries: .
– Using ATen parallel backend: OMP
– Using Lib/site-packages as python relative installation path

– ******** Summary ********
– General:
– CMake version : 3.16.4
– CMake command : C:/Program Files/CMake/bin/cmake.exe
– System : Windows
– C++ compiler : C:/Program Files (x86)/Microsoft Visual Studio/2019/Community/VC/Tools/MSVC/14.24.28314/bin/Hostx64/x64/cl.exe
– C++ compiler id : MSVC
– C++ compiler version : 19.24.28316.0
– BLAS : MKL
– CXX flags : /DWIN32 /D_WINDOWS /GR /w /EHa /MP /bigobj -openmp:experimental
– Build type : Release
– Compile definitions : ONNX_ML=1;ONNX_NAMESPACE=onnx_torch;_CRT_SECURE_NO_DEPRECATE=1;WIN32_LEAN_AND_MEAN
– CMAKE_PREFIX_PATH : C:\PtModelEnv\anaconda\envs\env_pytorch1.4\Lib\site-packages;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1
– CMAKE_INSTALL_PREFIX : C:/PtModelEnv/libs/pytorch/pytorch/torch

– TORCH_VERSION : 1.4.0
– CAFFE2_VERSION : 1.4.0
– BUILD_CAFFE2_MOBILE : ON
– USE_STATIC_DISPATCH : OFF
– BUILD_BINARY : OFF
– BUILD_CUSTOM_PROTOBUF : ON
– Link local protobuf : ON
– BUILD_DOCS : OFF
– BUILD_PYTHON : True
– Python version : 3.7.6
– Python executable : C:/PtModelEnv/anaconda/envs/env_pytorch1.4/python.exe
– Pythonlibs version : 3.7.6
– Python library : C:/PtModelEnv/anaconda/envs/env_pytorch1.4/libs/python37.lib
– Python includes : C:/PtModelEnv/anaconda/envs/env_pytorch1.4/include
– Python site-packages: Lib/site-packages
– BUILD_CAFFE2_OPS : OFF
– BUILD_SHARED_LIBS : ON
– BUILD_TEST : True
– BUILD_JNI : OFF
– INTERN_BUILD_MOBILE :
– USE_ASAN : OFF
– USE_CUDA : ON
– CUDA static link : OFF
– USE_CUDNN : OFF
– CUDA version : 10.1
– CUDA root directory : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1
– CUDA library : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/lib/x64/cuda.lib
– cudart library : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/lib/x64/cudart_static.lib
– cublas library : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/lib/x64/cublas.lib
– cufft library : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/lib/x64/cufft.lib
– curand library : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/lib/x64/curand.lib
– nvrtc : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/lib/x64/nvrtc.lib
– CUDA include path : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/include
– NVCC executable : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/bin/nvcc.exe
– CUDA host compiler : C:/Program Files (x86)/Microsoft Visual Studio/2019/Community/VC/Tools/MSVC/14.24.28314/bin/Hostx64/x64/cl.exe
– USE_TENSORRT : OFF
– USE_ROCM : OFF
– USE_EIGEN_FOR_BLAS : ON
– USE_FBGEMM : OFF
– USE_FFMPEG : OFF
– USE_GFLAGS : OFF
– USE_GLOG : OFF
– USE_LEVELDB : OFF
– USE_LITE_PROTO : OFF
– USE_LMDB : OFF
– USE_METAL : OFF
– USE_MKL : OFF
– USE_MKLDNN : OFF
– USE_NCCL : OFF
– USE_NNPACK : OFF
– USE_NUMPY : ON
– USE_OBSERVERS : ON
– USE_OPENCL : OFF
– USE_OPENCV : OFF
– USE_OPENMP : ON
– USE_TBB : OFF
– USE_PROF : OFF
– USE_QNNPACK : OFF
– USE_REDIS : OFF
– USE_ROCKSDB : OFF
– USE_ZMQ : OFF
– USE_DISTRIBUTED : OFF
– BUILD_NAMEDTENSOR : OFF
– Public Dependencies : Threads::Threads
– Private Dependencies : cpuinfo;fp16;foxi_loader
– Configuring done
– Generating done
– Build files have been written to: C:/PtModelEnv/libs/pytorch/pytorch/build
[1/2208] Building CXX object third_party\protobuf\cmake\CMakeFiles\libprotoc.dir__\src\google\protobuf\compiler\js\well_known_types_embed.cc.obj

.
.
.
[1492/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\quantized\cpu\kernels\QuantizedOpKernels.cpp.AVX.cpp.obj
[1493/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\TH\THTensorRandom.cpp.obj
[1494/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\TypeDefault.cpp.obj
[1495/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\layer_norm_kernel.cpp.DEFAULT.cpp.obj
[1496/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\DistanceOpsKernel.cpp.AVX.cpp.obj
[1497/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\PointwiseOpsKernel.cpp.AVX2.cpp.obj
[1498/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\Activation.cpp.AVX.cpp.obj
[1499/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\TensorCompareKernel.cpp.DEFAULT.cpp.obj
[1500/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\CopyKernel.cpp.AVX2.cpp.obj
[1501/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\event.cc.obj
[1502/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\FillKernel.cpp.AVX.cpp.obj
[1503/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\context_base.cc.obj
[1504/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\context.cc.obj
[1505/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\TH\THTensorMoreMath.cpp.obj
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(1013) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(1014) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(1013) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(1014) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(987) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(998) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(987) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(994) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(987) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(994) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(987) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(994) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(987) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorMoreMath.cpp(990) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
[1506/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\CopyKernel.cpp.AVX.cpp.obj
[1507/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\SortingKernel.cpp.DEFAULT.cpp.obj
[1508/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\PowKernel.cpp.DEFAULT.cpp.obj
[1509/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\SoftMaxKernel.cpp.DEFAULT.cpp.obj
[1510/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\CopyKernel.cpp.DEFAULT.cpp.obj
[1511/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\IndexKernel.cpp.DEFAULT.cpp.obj
[1512/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\CrossKernel.cpp.DEFAULT.cpp.obj
[1513/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\BinaryOpsKernel.cpp.AVX.cpp.obj
[1514/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\DepthwiseConvKernel.cpp.DEFAULT.cpp.obj
[1515/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\TH\THTensorEvenMoreMath.cpp.obj
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(846) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(882) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(329) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(778) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(816) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(848) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(329) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(780) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(818) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(848) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(329) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(780) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(818) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(848) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(329) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(780) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(818) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(848) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(329) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(780) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(818) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(848) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(329) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(197) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(882) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(846) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(886) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(886) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(886) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(886) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(886) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(279) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
C:\PtModelEnv\libs\pytorch\pytorch\aten\src\TH\generic\THTensorEvenMoreMath.cpp(292) : info C5002: Omp simd loop not vectorized due to reason ‘1200’
[1516/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\Unfold2d.cpp.DEFAULT.cpp.obj
[1517/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\PointwiseOpsKernel.cpp.DEFAULT.cpp.obj
[1518/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\UnaryOpsKernel.cpp.AVX2.cpp.obj
[1519/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\DistanceOpsKernel.cpp.DEFAULT.cpp.obj
[1520/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\LerpKernel.cpp.DEFAULT.cpp.obj
[1521/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\ReduceOpsKernel.cpp.DEFAULT.cpp.obj
[1522/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\MultinomialKernel.cpp.DEFAULT.cpp.obj
[1523/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\FillKernel.cpp.DEFAULT.cpp.obj
[1524/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\graph.cc.obj
[1525/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\blob_stats.cc.obj
[1526/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\quantized\cpu\kernels\QuantizedOpKernels.cpp.DEFAULT.cpp.obj
[1527/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\Activation.cpp.DEFAULT.cpp.obj
[1528/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\UnaryOpsKernel.cpp.DEFAULT.cpp.obj
[1529/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\db.cc.obj
[1530/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\GridSamplerKernel.cpp.DEFAULT.cpp.obj
[1531/2208] Building CXX object caffe2\CMakeFiles\torch.dir\core\export_c10_op_to_caffe2.cc.obj
[1532/2208] Building CXX object caffe2\CMakeFiles\torch.dir_\aten\src\ATen\native\cpu\BinaryOpsKernel.cpp.DEFAULT.cpp.obj
ninja: build stopped: subcommand failed.
– Building version 1.4.0a0+7f73f1d
cmake -GNinja -DBUILD_PYTHON=True -DBUILD_TEST=True -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=C:\PtModelEnv\libs\pytorch\pytorch\torch -DCMAKE_PREFIX_PATH=C:\PtModelEnv\anaconda\envs\env_pytorch1.4\Lib\site-packages -DNUMPY_INCLUDE_DIR=C:\PtModelEnv\anaconda\envs\env_pytorch1.4\lib\site-packages\numpy\core\include -DPYTHON_EXECUTABLE=C:\PtModelEnv\anaconda\envs\env_pytorch1.4\python.exe -DPYTHON_INCLUDE_DIR=C:\PtModelEnv\anaconda\envs\env_pytorch1.4\include -DPYTHON_LIBRARY=C:\PtModelEnv\anaconda\envs\env_pytorch1.4/libs/python37.lib -DTORCH_BUILD_VERSION=1.4.0a0+7f73f1d -DUSE_NUMPY=True C:\PtModelEnv\libs\pytorch\pytorch
cmake --build . --target install --config Release – -j 48

5

Also, when is pytorch1.5 Windows planned for release?

6

python setup.py build 2>&1 > full_log.txt

7

But I can see those DLLs in the nightly package. Maybe you are doing sth wrong. Could you please tell me how did you build it?

8

Well, the PR that brings MKLDNN support for Windows is coming after the release of 1.4.0. So they aren’t supported.

9

The log appears too large to paste in it’s entirety and only image files (jpg, png, etc) can be uploaded.

10

And another question - I wanted to find out if any plans to release pytorch1.5 in the next 1-1.5 months?

11

@smth may know the answer.

12

Use sth like https://gist.github.com/ or https://pastebin.com/.

13

1.5 is going to be in April.