sync w/ upstream

.gitignore

@@ -8,58 +8,59 @@
 
 ## PyTorch
 
-build/
-dist/
-torch.egg-info/
+.mypy_cache
+*/*.pyc
+*/*.so*
 */**/__pycache__
+*/**/*.dylib*
+*/**/*.pyc
+*/**/*.pyd
+*/**/*.so*
+*/**/**/*.pyc
+*/**/**/**/*.pyc
+*/**/**/**/**/*.pyc
 aten/build/
 aten/src/ATen/Config.h
 aten/src/ATen/cuda/CUDAConfig.h
+build/
+dist/
+docs/src/**/*
+test/.coverage
+test/cpp/api/mnist
+test/data/gpu_tensors.pt
+test/data/legacy_modules.t7
+test/data/legacy_serialized.pt
+test/data/linear.pt
+test/htmlcov
 third_party/build/
-torch/version.py
+tools/shared/_utils_internal.py
+torch.egg-info/
+torch/csrc/autograd/generated/*
+torch/csrc/cudnn/cuDNN.cpp
+torch/csrc/generated
 torch/csrc/generic/TensorMethods.cpp
-torch/lib/*.so*
+torch/csrc/jit/generated/*
+torch/csrc/nn/THCUNN.cpp
+torch/csrc/nn/THCUNN.cwrap
+torch/csrc/nn/THNN_generic.cpp
+torch/csrc/nn/THNN_generic.cwrap
+torch/csrc/nn/THNN_generic.h
+torch/csrc/nn/THNN.cpp
+torch/csrc/nn/THNN.cwrap
 torch/lib/*.a*
 torch/lib/*.dll*
-torch/lib/*.lib
 torch/lib/*.dylib*
 torch/lib/*.h
+torch/lib/*.lib
+torch/lib/*.so*
 torch/lib/build
 torch/lib/cmake
+torch/lib/include
 torch/lib/pkgconfig
 torch/lib/protoc
 torch/lib/tmp_install
-torch/lib/include
 torch/lib/torch_shm_manager
-torch/csrc/jit/generated/*
-torch/csrc/autograd/generated/*
-torch/csrc/cudnn/cuDNN.cpp
-torch/csrc/nn/THNN.cwrap
-torch/csrc/nn/THNN.cpp
-torch/csrc/nn/THCUNN.cwrap
-torch/csrc/nn/THCUNN.cpp
-torch/csrc/nn/THNN_generic.cwrap
-torch/csrc/nn/THNN_generic.cpp
-torch/csrc/nn/THNN_generic.h
-torch/csrc/generated
-docs/src/**/*
-test/data/legacy_modules.t7
-test/data/gpu_tensors.pt
-test/htmlcov
-test/.coverage
-*/*.pyc
-*/**/*.pyc
-*/**/**/*.pyc
-*/**/**/**/*.pyc
-*/**/**/**/**/*.pyc
-*/*.so*
-*/**/*.so*
-*/**/*.dylib*
-*/**/*.pyd
-test/data/legacy_serialized.pt
-test/data/linear.pt
-.mypy_cache
-test/cpp/api/mnist
+torch/version.py
 
 # IPython notebook checkpoints
 .ipynb_checkpoints

.gitmodules

@@ -1,7 +1,3 @@
-[submodule "third_party/tbb"]
-	path = third_party/tbb
-	url = https://github.com/01org/tbb
-	branch = tbb_2018
 [submodule "third_party/catch"]
 	path = third_party/catch
 	url = https://github.com/catchorg/Catch2.git

.jenkins/pytorch/build.sh

@@ -4,8 +4,11 @@ if [[ "$BUILD_ENVIRONMENT" == "pytorch-linux-xenial-py3-clang5-asan" ]]; then
   exec "$(dirname "${BASH_SOURCE[0]}")/build-asan.sh" $*
 fi
 
-# Add nccl2 for distributed test.
-apt-get install libnccl-dev libnccl2
+# TODO: move this to Docker
+# TODO: add both NCCL and MPI in CI test by fixing these test first 
+# sudo apt-get update
+# sudo apt-get install libnccl-dev libnccl2
+# sudo apt-get install openmpi-bin libopenmpi-dev
 
 # Required environment variable: $BUILD_ENVIRONMENT
 # (This is set by default in the Docker images we build, so you don't
@@ -36,7 +39,7 @@ if [[ "$BUILD_ENVIRONMENT" == *rocm* ]]; then
   sudo chown -R jenkins:jenkins /usr/local
   rm -rf "$(dirname "${BASH_SOURCE[0]}")/../../../pytorch_amd/" || true
   python "$(dirname "${BASH_SOURCE[0]}")/../../tools/amd_build/build_pytorch_amd.py"
-  HIPCC_VERBOSE=7 VERBOSE=1 WITH_ROCM=1 python setup.py install
+  USE_ROCM=1 python setup.py install
   exit
 fi
 
@@ -46,14 +49,18 @@ if ! which conda; then
 fi
 
 # sccache will fail for CUDA builds if all cores are used for compiling
-# gcc 7.2 with sccache seems to have intermittent OOM issue if all cores are used
-if ([[ "$BUILD_ENVIRONMENT" == *cuda* ]] || [[ "$BUILD_ENVIRONMENT" == *gcc7.2* ]]) && which sccache > /dev/null; then
+# gcc 7 with sccache seems to have intermittent OOM issue if all cores are used
+if ([[ "$BUILD_ENVIRONMENT" == *cuda* ]] || [[ "$BUILD_ENVIRONMENT" == *gcc7* ]]) && which sccache > /dev/null; then
   export MAX_JOBS=`expr $(nproc) - 1`
 fi
 
 # Target only our CI GPU machine's CUDA arch to speed up the build
 export TORCH_CUDA_ARCH_LIST=5.2
 
+if [[ "$BUILD_ENVIRONMENT" == *trusty-py3.6-gcc5.4* ]]; then
+  export DEBUG=1
+fi
+
 WERROR=1 python setup.py install
 
 # Add the test binaries so that they won't be git clean'ed away

.jenkins/pytorch/common.sh

@@ -113,7 +113,7 @@ else
 fi
 
 if [[ "$BUILD_ENVIRONMENT" == *pytorch-linux-xenial-cuda9-cudnn7-py3 ]] || \
-   [[ "$BUILD_ENVIRONMENT" == *pytorch-linux-trusty-py3.6-gcc7.2 ]]; then
+   [[ "$BUILD_ENVIRONMENT" == *pytorch-linux-trusty-py3.6-gcc7* ]]; then
   BUILD_TEST_LIBTORCH=1
 else
   BUILD_TEST_LIBTORCH=0

.jenkins/pytorch/enabled-configs.txt

@@ -12,6 +12,8 @@ pytorch-linux-xenial-cuda9-cudnn7-py2-build
 pytorch-linux-xenial-cuda9-cudnn7-py2-test
 pytorch-linux-xenial-cuda9-cudnn7-py3-build
 pytorch-linux-xenial-cuda9-cudnn7-py3-test
+pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc7-build
+pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc7-test
 pytorch-linux-xenial-py3-clang5-asan-build
 pytorch-linux-xenial-py3-clang5-asan-test
 pytorch-linux-trusty-py2.7.9-build
@@ -26,6 +28,8 @@ pytorch-linux-trusty-py3.6-gcc5.4-build
 pytorch-linux-trusty-py3.6-gcc5.4-test
 pytorch-linux-trusty-py3.6-gcc7.2-build
 pytorch-linux-trusty-py3.6-gcc7.2-test
+pytorch-linux-trusty-py3.6-gcc7-build
+pytorch-linux-trusty-py3.6-gcc7-test
 pytorch-linux-trusty-pynightly-build
 pytorch-linux-trusty-pynightly-test
 pytorch-win-ws2016-cuda9-cudnn7-py3-build

.jenkins/pytorch/win-build.sh

@@ -1,6 +1,13 @@
 #!/bin/bash
 
 # If you want to rebuild, run this with REBUILD=1
+# If you want to build with CUDA, run this with USE_CUDA=1
+# If you want to build without CUDA, run this with USE_CUDA=0
+
+if [ ! -f setup.py ]; then
+  echo "ERROR: Please run this build script from PyTorch root directory."
+  exit 1
+fi
 
 COMPACT_JOB_NAME=pytorch-win-ws2016-cuda9-cudnn7-py3-build
 source "$(dirname "${BASH_SOURCE[0]}")/common.sh"
@@ -34,12 +41,26 @@ cat >ci_scripts/build_pytorch.bat <<EOL
 set PATH=C:\\Program Files\\CMake\\bin;C:\\Program Files\\7-Zip;C:\\curl-7.57.0-win64-mingw\\bin;C:\\Program Files\\Git\\cmd;C:\\Program Files\\Amazon\\AWSCLI;%PATH%
 
 :: Install MKL
-if "%REBUILD%"=="" ( aws s3 cp s3://ossci-windows/mkl_2018.2.185.7z mkl.7z --quiet && 7z x -aoa mkl.7z -omkl )
+if "%REBUILD%"=="" (
+  if "%BUILD_ENVIRONMENT%"=="" (
+    curl -k https://s3.amazonaws.com/ossci-windows/mkl_2018.2.185.7z --output mkl.7z
+  ) else (
+    aws s3 cp s3://ossci-windows/mkl_2018.2.185.7z mkl.7z --quiet
+  )
+  7z x -aoa mkl.7z -omkl
+)
 set CMAKE_INCLUDE_PATH=%cd%\\mkl\\include
 set LIB=%cd%\\mkl\\lib;%LIB
 
 :: Install MAGMA
-if "%REBUILD%"=="" ( aws s3 cp s3://ossci-windows/magma_cuda90_release_mkl_2018.2.185.7z magma_cuda90_release_mkl_2018.2.185.7z --quiet && 7z x -aoa magma_cuda90_release_mkl_2018.2.185.7z -omagma )
+if "%REBUILD%"=="" (
+  if "%BUILD_ENVIRONMENT%"=="" (
+    curl -k https://s3.amazonaws.com/ossci-windows/magma_cuda90_release_mkl_2018.2.185.7z --output magma_cuda90_release_mkl_2018.2.185.7z
+  ) else (
+    aws s3 cp s3://ossci-windows/magma_cuda90_release_mkl_2018.2.185.7z magma_cuda90_release_mkl_2018.2.185.7z --quiet
+  )
+  7z x -aoa magma_cuda90_release_mkl_2018.2.185.7z -omagma
+)
 set MAGMA_HOME=%cd%\\magma
 
 :: Install sccache
@@ -49,15 +70,19 @@ if "%REBUILD%"=="" (
   %CD%\\tmp_bin\\sccache.exe --show-stats || (
     taskkill /im sccache.exe /f /t || ver > nul
     del %CD%\\tmp_bin\\sccache.exe
-    aws s3 cp s3://ossci-windows/sccache.exe %CD%\\tmp_bin\\sccache.exe
+    if "%BUILD_ENVIRONMENT%"=="" (
+      curl -k https://s3.amazonaws.com/ossci-windows/sccache.exe --output %CD%\\tmp_bin\\sccache.exe
+    ) else (
+      aws s3 cp s3://ossci-windows/sccache.exe %CD%\\tmp_bin\\sccache.exe
+    )
     goto :check_sccache
   )
 )
 
 :: Install Miniconda3
 if "%REBUILD%"=="" (
   IF EXIST C:\\Jenkins\\Miniconda3 ( rd /s /q C:\\Jenkins\\Miniconda3 )
-  curl https://repo.continuum.io/miniconda/Miniconda3-latest-Windows-x86_64.exe -O
+  curl -k https://repo.continuum.io/miniconda/Miniconda3-latest-Windows-x86_64.exe -O
   .\Miniconda3-latest-Windows-x86_64.exe /InstallationType=JustMe /RegisterPython=0 /S /AddToPath=0 /D=C:\\Jenkins\\Miniconda3
 )
 call C:\\Jenkins\\Miniconda3\\Scripts\\activate.bat C:\\Jenkins\\Miniconda3
@@ -91,24 +116,35 @@ set DISTUTILS_USE_SDK=1
 
 set CMAKE_GENERATOR=Ninja
 
-if "%REBUILD%"=="" (
-  set NO_CUDA=1
-  python setup.py install
-)
-if errorlevel 1 exit /b 1
-if not errorlevel 0 exit /b 1
-if "%REBUILD%"=="" (
-  sccache --show-stats
-  sccache --zero-stats
-  rd /s /q C:\\Jenkins\\Miniconda3\\Lib\\site-packages\\torch
-  copy %CD%\\tmp_bin\\sccache.exe tmp_bin\\nvcc.exe
+if not "%USE_CUDA%"=="1" (
+  if "%REBUILD%"=="" (
+    set NO_CUDA=1
+    python setup.py install
+  )
+  if errorlevel 1 exit /b 1
+  if not errorlevel 0 exit /b 1
 )
 
-set CUDA_NVCC_EXECUTABLE=%CD%\\tmp_bin\\nvcc
+if not "%USE_CUDA%"=="0" (
+  if "%REBUILD%"=="" (
+    sccache --show-stats
+    sccache --zero-stats
+    rd /s /q C:\\Jenkins\\Miniconda3\\Lib\\site-packages\\torch
+    copy %CD%\\tmp_bin\\sccache.exe tmp_bin\\nvcc.exe
+  )
+
+  set CUDA_NVCC_EXECUTABLE=%CD%\\tmp_bin\\nvcc
 
-if "%REBUILD%"=="" set NO_CUDA=
+  if "%REBUILD%"=="" set NO_CUDA=0
 
-python setup.py install && sccache --show-stats && 7z a %IMAGE_COMMIT_TAG%.7z C:\\Jenkins\\Miniconda3\\Lib\\site-packages\\torch && python ci_scripts\\upload_image.py %IMAGE_COMMIT_TAG%.7z
+  python setup.py install && sccache --show-stats && (
+    if "%BUILD_ENVIRONMENT%"=="" (
+      echo "NOTE: To run `import torch`, please make sure to activate the conda environment by running `call C:\\Jenkins\\Miniconda3\\Scripts\\activate.bat C:\\Jenkins\\Miniconda3` in Command Prompt before running Git Bash."
+    ) else (
+      7z a %IMAGE_COMMIT_TAG%.7z C:\\Jenkins\\Miniconda3\\Lib\\site-packages\\torch && python ci_scripts\\upload_image.py %IMAGE_COMMIT_TAG%.7z
+    )
+  )
+)
 
 EOL
 

CMakeLists.txt

@@ -138,31 +138,6 @@ cmake_dependent_option(
 option(USE_DISTRIBUTED "Use THD (distributed)" OFF)
 option(USE_DISTRIBUTED_MW "Use THD (distributed) master worker" OFF)
 
-# Legacy options, which we will eventually remove
-cmake_dependent_option(
-    WITH_CUDA "Legacy CUDA" ON
-    "USE_CUDA" OFF)
-cmake_dependent_option(
-    NO_PYTHON "Legacy Python" OFF
-    "BUILD_PYTHON" ON)
-cmake_dependent_option(
-    WITH_CUDNN "Legacy cuDNN" ON
-    "USE_CUDNN" OFF)
-cmake_dependent_option(
-    WITH_NCCL "Legacy NCCL" ON
-    "USE_NCCL" OFF)
-cmake_dependent_option(
-    NO_MKLDNN "Legacy no MKLDNN" OFF
-    "USE_MKLDNN" ON)
-cmake_dependent_option(
-    WITH_DISTRIBUTED "Legacy THD (distributed)" ON
-    "USE_DISTRIBUTED" OFF)
-cmake_dependent_option(
-    WITH_DISTRIBUTED_MW "Legacy THD (distributed) MW" ON
-    "USE_DISTRIBUTED_MW" OFF)
-cmake_dependent_option(
-    WITH_GLOO_IBVERBS "Legacy Gloo IB verbs for distributed support" ON
-    "USE_GLOO_IBVERBS" OFF)
 if (USE_ATEN)
     set(BUILD_ATEN ${USE_ATEN})
 endif()

CODEOWNERS

@@ -3,7 +3,7 @@
 
 /aten/ @apaszke @soumith @colesbury @gchanan @zdevito @ezyang
 /torch/ @apaszke @soumith @colesbury @gchanan @zdevito @ezyang
-/docs/source @apaszke @soumith @colesbury @gchanan @zdevito @ezyang
+/docs/source @apaszke @soumith @colesbury @gchanan @zdevito @ezyang @ssnl @zou3519
 /test @apaszke @soumith @colesbury @gchanan @zdevito @ezyang
 /tools @apaszke @soumith @colesbury @gchanan @zdevito @ezyang
 /README.md @apaszke @soumith @colesbury @gchanan @zdevito @ezyang
@@ -20,3 +20,5 @@
 /torch/csrc/distributed/ @apaszke @pietern @teng-li
 /torch/distributed/ @apaszke @pietern @teng-li
 /test/test_c10d.py @apaszke @pietern @teng-li
+/third_party/ @orionr
+/torch/utils/cpp_extension.py @goldsborough @fmassa @apaszke @soumith @ezyang

README.md

@@ -141,9 +141,8 @@ and with minimal abstractions.
 You can write new neural network layers in Python using the torch API
 [or your favorite NumPy-based libraries such as SciPy](http://pytorch.org/tutorials/advanced/numpy_extensions_tutorial.html).
 
-If you want to write your layers in C/C++, we provide an extension API based on
-[cffi](http://cffi.readthedocs.io/en/latest/) that is efficient and with minimal boilerplate.
-There is no wrapper code that needs to be written. You can see [a tutorial here](http://pytorch.org/tutorials/advanced/c_extension.html) and [an example here](https://github.com/pytorch/extension-ffi).
+If you want to write your layers in C/C++, we provide a convenient extension API that is efficient and with minimal boilerplate.
+There is no wrapper code that needs to be written. You can see [a tutorial here](http://pytorch.org/tutorials/advanced/cpp_extension.html) and [an example here](https://github.com/pytorch/extension-cpp).
 
 
 ## Installation

aten/CMakeLists.txt

@@ -4,6 +4,7 @@ if (CAFFE2_CMAKE_BUILDING_WITH_MAIN_REPO)
   endif()
 else()
   cmake_minimum_required(VERSION 3.0 FATAL_ERROR)
+  project(ATen CXX C)
   include(CMakeDependentOption)
   option(USE_CUDA "Use CUDA" ON)
   option(USE_ROCM "Use ROCm" OFF)
@@ -14,21 +15,10 @@ else()
       "USE_CUDA" OFF)
   option(ATEN_NO_TEST "Do not build ATen test binaries" OFF)
 
-  # Legacy options, which we will eventually remove
-  cmake_dependent_option(
-      WITH_CUDNN "Legacy cuDNN" ON
-      "USE_CUDNN" OFF)
-  cmake_dependent_option(
-      NO_MKLDNN "Legacy no MKLDNN" OFF
-      "USE_MKLDNN" ON)
-
   # Flag for shared dependencies
   set(BUILD_ATEN ON)
 endif()
 
-# Create the project in all cases
-project(ATen CXX C)
-
 # Find modules
 list(APPEND CMAKE_MODULE_PATH
   /usr/lib/x86_64-linux-gnu/
@@ -39,9 +29,6 @@ list(APPEND CMAKE_LIBRARY_PATH /usr/lib/x86_64-linux-gnu/)
 
 cmake_policy(SET CMP0012 NEW)
 
-# Polyfill for upstream FindCUDA
-include(CMakeInitializeConfigs)
-
 #############################################
 
 set(ATen_CPU_SRCS)