Merge pull request #14 from iotamudelta/master

merge from upstream

Author: iotamudelta

.gitattributes

@@ -90,20 +90,24 @@ if [[ $BUILD_ENVIRONMENT == conda* ]]; then
   conda_ignore_test+=("--ignore $CAFFE2_PYPATH/python/operator_test/checkpoint_test.py")
 fi
 
-# Python tests
-echo "Running Python tests.."
-"$PYTHON" \
-  -m pytest \
-  -x \
-  -v \
-  --junit-xml="$TEST_DIR/python/result.xml" \
-  --ignore "$CAFFE2_PYPATH/python/test/executor_test.py" \
-  --ignore "$CAFFE2_PYPATH/python/operator_test/matmul_op_test.py" \
-  --ignore "$CAFFE2_PYPATH/python/operator_test/pack_ops_test.py" \
-  --ignore "$CAFFE2_PYPATH/python/mkl/mkl_sbn_speed_test.py" \
-  ${conda_ignore_test[@]} \
-  "$CAFFE2_PYPATH/python" \
-  "${EXTRA_TESTS[@]}"
+
+# TODO: re-enable this for rocm CI jobs once we have more rocm workers
+if [[ $BUILD_ENVIRONMENT != *rocm* ]]; then
+  # Python tests
+  echo "Running Python tests.."
+  "$PYTHON" \
+    -m pytest \
+    -x \
+    -v \
+    --junit-xml="$TEST_DIR/python/result.xml" \
+    --ignore "$CAFFE2_PYPATH/python/test/executor_test.py" \
+    --ignore "$CAFFE2_PYPATH/python/operator_test/matmul_op_test.py" \
+    --ignore "$CAFFE2_PYPATH/python/operator_test/pack_ops_test.py" \
+    --ignore "$CAFFE2_PYPATH/python/mkl/mkl_sbn_speed_test.py" \
+    ${conda_ignore_test[@]} \
+    "$CAFFE2_PYPATH/python" \
+    "${EXTRA_TESTS[@]}"
+fi
 
 if [[ -n "$INTEGRATED" ]]; then
   pip install --user pytest-xdist torchvision

.jenkins/caffe2/test.sh

@@ -5,7 +5,7 @@ if [[ "$BUILD_ENVIRONMENT" == "pytorch-linux-xenial-py3-clang5-asan" ]]; then
 fi
 
 # TODO: move this to Docker
-# TODO: add both NCCL and MPI in CI test by fixing these test first 
+# 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
@@ -23,6 +23,9 @@ python --version
 echo "GCC version:"
 gcc --version
 
+echo "CMake version:"
+cmake --version
+
 # TODO: Don't run this...
 pip install -r requirements.txt || true
 

.jenkins/pytorch/build.sh

@@ -118,3 +118,23 @@ if [[ "$BUILD_ENVIRONMENT" == *pytorch-linux-xenial-cuda9-cudnn7-py3 ]] || \
 else
   BUILD_TEST_LIBTORCH=0
 fi
+
+# Use conda cmake in some CI build. Conda cmake will be newer than our supported
+# min version 3.5, so we only do it in two builds that we know should use conda.
+if [[ "$BUILD_ENVIRONMENT" == *pytorch-linux-xenial-cuda* ]]; then
+  if [[ "$BUILD_ENVIRONMENT" == *cuda8-cudnn6-py2* ]] || \
+     [[ "$BUILD_ENVIRONMENT" == *cuda9-cudnn7-py3* ]]; then
+    if ! which conda; then
+      echo "Expected ${BUILD_ENVIRONMENT} to use conda, but 'which conda' returns empty"
+      exit 1
+    else
+      conda install -q -y cmake
+    fi
+  else
+    if ! cmake --version | grep 'cmake version 3\.5'; then
+      echo "Expected ${BUILD_ENVIRONMENT} to have cmake version 3.5.* (min support version), but 'cmake --version' returns:"
+      cmake --version
+      exit 1
+    fi
+  fi
+fi

.jenkins/pytorch/common.sh

@@ -11,7 +11,7 @@ get_runtime_of_command () {
 
   # runtime=$( { time ($@ &> /dev/null); } 2>&1 1>/dev/null)
   runtime=$( { time $@; } 2>&1 1>/dev/null)
-  if [[ $runtime == *"Warning"* ]] || [[ $runtime == *"Error"* ]]; then
+  if [[ $runtime == *"Error"* ]]; then
     exit 1
   fi
   runtime=${runtime#+++ $@}

.jenkins/pytorch/perf_test/common.sh

@@ -139,7 +139,7 @@ if not "%USE_CUDA%"=="0" (
 
   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."
+      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
     )
@@ -148,4 +148,8 @@ if not "%USE_CUDA%"=="0" (
 
 EOL
 
-ci_scripts/build_pytorch.bat && echo "BUILD PASSED"
+ci_scripts/build_pytorch.bat
+if [ ! -f $IMAGE_COMMIT_TAG.7z ] && [ ! ${BUILD_ENVIRONMENT} == "" ]; then
+    exit 1
+fi
+echo "BUILD PASSED"

.jenkins/pytorch/win-build.sh

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.2 FATAL_ERROR)
+cmake_minimum_required(VERSION 3.5 FATAL_ERROR)
 #cmake_policy(SET CMP0022 NEW)
 #cmake_policy(SET CMP0023 NEW)
 
@@ -197,23 +197,27 @@ if(NOT MSVC)
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing")
   # Eigen fails to build with some versions, so convert this to a warning
   # Details at http://eigen.tuxfamily.org/bz/show_bug.cgi?id=1459
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-invalid-partial-specialization")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wextra")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-missing-field-initializers")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-type-limits")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-typedef-redefinition")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unknown-warning-option")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unknown-pragmas")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-sign-compare")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-parameter")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-variable")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-function")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-private-field")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-result")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-inconsistent-missing-override")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-aligned-allocation-unavailable")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-error=deprecated-declarations")
+  # These flags are not available in GCC-4.8.5. Set only when using clang.
+  # Compared against https://gcc.gnu.org/onlinedocs/gcc-4.8.5/gcc/Option-Summary.html
+  if ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-invalid-partial-specialization")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-typedef-redefinition")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unknown-warning-option")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-private-field")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-inconsistent-missing-override")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-aligned-allocation-unavailable")
+  endif()
   if ((APPLE AND (NOT ("${CLANG_VERSION_STRING}" VERSION_LESS "9.0")))
     OR (CMAKE_COMPILER_IS_GNUCXX AND (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 7.0)))
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -faligned-new")

CMakeLists.txt

@@ -20,5 +20,4 @@
 /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

CODEOWNERS

@@ -27,7 +27,7 @@ We are in an early-release beta. Expect some adventures and rough edges.
 | Linux GPU | [![Build Status](https://ci.pytorch.org/jenkins/job/pytorch-master/badge/icon)](https://ci.pytorch.org/jenkins/job/pytorch-master/) | [![Build Status](https://ci.pytorch.org/jenkins/job/pytorch-master/badge/icon)](https://ci.pytorch.org/jenkins/job/pytorch-master/) |
 | Windows GPU | <center>—</center> | [![Build Status](https://ci.pytorch.org/jenkins/job/pytorch-builds/job/pytorch-win-ws2016-cuda9-cudnn7-py3-trigger/badge/icon)](https://ci.pytorch.org/jenkins/job/pytorch-builds/job/pytorch-win-ws2016-cuda9-cudnn7-py3-trigger/)
 
-See also the [ci.pytorch.org HUD](https://ezyang.github.io/pytorch-ci-hud/build/pytorch-master)
+See also the [ci.pytorch.org HUD](https://ezyang.github.io/pytorch-ci-hud/build/pytorch-master).
 
 
 ## More about PyTorch

README.md

@@ -83,18 +83,12 @@ set(TH_CPU_INCLUDE
   ${CMAKE_CURRENT_SOURCE_DIR}/src/THC
   ${CMAKE_CURRENT_BINARY_DIR}/src/TH
   ${CMAKE_CURRENT_BINARY_DIR}/src/THC
-  # sparse
-  ${CMAKE_CURRENT_SOURCE_DIR}/src/THS
-  ${CMAKE_CURRENT_SOURCE_DIR}/src/THCS
-  ${CMAKE_CURRENT_BINARY_DIR}/src/THS
-  ${CMAKE_CURRENT_BINARY_DIR}/src/THCS
 
   ${CMAKE_CURRENT_SOURCE_DIR}/src
   ${CMAKE_CURRENT_BINARY_DIR}/src
   ${CMAKE_BINARY_DIR}/aten/src)
 list(APPEND ATen_CPU_INCLUDE ${TH_CPU_INCLUDE})
 add_subdirectory(src/THNN)
-add_subdirectory(src/THS)
 
 # Find the HIP package, set the HIP paths, load the HIP CMake.
 IF(USE_ROCM)
@@ -115,13 +109,11 @@ if(USE_ROCM)
   SET(AT_CUDA_ENABLED 1)
   add_subdirectory(src/THC)
   add_subdirectory(src/THCUNN)
-  add_subdirectory(src/THCS)
   message("ROCm is enabled.")
 elseif(USE_CUDA)
   SET(AT_CUDA_ENABLED 1)
   add_subdirectory(src/THC)
   add_subdirectory(src/THCUNN)
-  add_subdirectory(src/THCS)
 else()
   message("disabling CUDA because USE_CUDA is set false")
   SET(AT_CUDA_ENABLED 0)

aten/CMakeLists.txt

@@ -20,3 +20,4 @@
 #include "ATen/DeviceGuard.h"
 #include "ATen/TensorOptions.h"
 #include "ATen/Layout.h"
+#include "ATen/OptionsGuard.h"

aten/src/ATen/ATen.h

@@ -13,22 +13,4 @@ struct Allocator {
   virtual void deallocate(void* ptr) const = 0;
 };
 
-namespace detail {
-
-struct AllocatorRetainable : public Retainable {
-  AllocatorRetainable(std::unique_ptr<Allocator> allocator)
-    : allocator(std::move(allocator)) {}
-
-  void* allocate(size_t n) {
-    return allocator->allocate(n);
-  }
-  void deallocate(void* ptr) {
-    return allocator->deallocate(ptr);
-  }
-private:
-  std::unique_ptr<Allocator> allocator;
-};
-
-}  // namespace at::detail
-
 }  // namespace at

aten/src/ATen/Allocator.h

@@ -12,13 +12,6 @@
 #include <mutex>
 #include <cstdint>
 
-// Forwarde declare these CUDA types here to avoid including CUDA headers in
-// ATen headers, which would make ATen always require CUDA to build.
-struct THCState;
-struct CUstream_st;
-typedef struct CUstream_st *cudaStream_t;
-struct cudaDeviceProp;
-
 namespace at {
 
 enum class IsVariable {
@@ -91,6 +84,11 @@ class AT_API Context {
   cudaStream_t getCurrentCUDAStreamOnDevice(int64_t device) const {
     return detail::getCUDAHooks().getCurrentCUDAStreamOnDevice(thc_state.get(), device);
   }
+#ifndef __HIP_PLATFORM_HCC__
+  cusparseHandle_t getCurrentCUDASparseHandle() const {
+    return detail::getCUDAHooks().getCurrentCUDASparseHandle(thc_state.get());
+  }
+#endif
   cudaDeviceProp* getCurrentDeviceProperties() const {
     return detail::getCUDAHooks().getCurrentDeviceProperties(thc_state.get());
   }