Revert D20589048: [pytorch][PR] [ROCm] Update CI dockers to ROCm rele…

…ase 3.1.1

Test Plan: revert-hammer

Differential Revision:
D20589048

Original commit changeset: 568f40c1b90f

fbshipit-source-id: 724c4fe99e8806f00d2f7dceb71d15a02358f663

.circleci/cimodel/data/caffe2_build_definitions.py

@@ -12,7 +12,7 @@
 
 DOCKER_IMAGE_PATH_BASE = "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/"
 
-DOCKER_IMAGE_VERSION = "369"
+DOCKER_IMAGE_VERSION = "345"
 
 
 @dataclass

.circleci/config.yml

@@ -2422,44 +2422,44 @@ workflows:
           requires:
             - setup
           build_environment: "caffe2-onnx-py2-gcc5-ubuntu16.04-build"
-          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py2-gcc5-ubuntu16.04:369"
+          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py2-gcc5-ubuntu16.04:345"
       - caffe2_linux_test:
           name: caffe2_onnx_py2_gcc5_ubuntu16_04_test
           requires:
             - setup
             - caffe2_onnx_py2_gcc5_ubuntu16_04_build
           build_environment: "caffe2-onnx-py2-gcc5-ubuntu16.04-test"
-          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py2-gcc5-ubuntu16.04:369"
+          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py2-gcc5-ubuntu16.04:345"
           resource_class: large
       - caffe2_linux_build:
           name: caffe2_onnx_main_py3_6_clang7_ubuntu16_04_build
           requires:
             - setup
           build_environment: "caffe2-onnx-main-py3.6-clang7-ubuntu16.04-build"
-          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:369"
+          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:345"
       - caffe2_linux_test:
           name: caffe2_onnx_main_py3_6_clang7_ubuntu16_04_test
           requires:
             - setup
             - caffe2_onnx_main_py3_6_clang7_ubuntu16_04_build
           build_environment: "caffe2-onnx-main-py3.6-clang7-ubuntu16.04-test"
-          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:369"
+          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:345"
           resource_class: large
       - caffe2_linux_test:
           name: caffe2_onnx_ort1_py3_6_clang7_ubuntu16_04_test
           requires:
             - setup
             - caffe2_onnx_main_py3_6_clang7_ubuntu16_04_build
           build_environment: "caffe2-onnx-ort1-py3.6-clang7-ubuntu16.04-test"
-          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:369"
+          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:345"
           resource_class: large
       - caffe2_linux_test:
           name: caffe2_onnx_ort2_py3_6_clang7_ubuntu16_04_test
           requires:
             - setup
             - caffe2_onnx_main_py3_6_clang7_ubuntu16_04_build
           build_environment: "caffe2-onnx-ort2-py3.6-clang7-ubuntu16.04-test"
-          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:369"
+          docker_image: "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/py3.6-clang7-ubuntu16.04:345"
           resource_class: large
       # TODO: Refactor circleci/cimodel/data/binary_build_data.py to generate this file
       #       instead of doing one offs here
@@ -6998,7 +6998,7 @@ workflows:
       - ecr_gc_job:
             name: ecr_gc_job_for_caffe2
             project: caffe2
-            tags_to_keep: "369,348,345,336,325,324,315,306,301,287,283,276,273,266,253,248,238,230,213"
+            tags_to_keep: "348,345,336,325,324,315,306,301,287,283,276,273,266,253,248,238,230,213"
       - ecr_gc_job:
             name: ecr_gc_job_for_translate
             project: translate

.circleci/docker/build.sh

@@ -15,8 +15,6 @@ OS="ubuntu"
 DOCKERFILE="${OS}/Dockerfile"
 if [[ "$image" == *-cuda* ]]; then
   DOCKERFILE="${OS}-cuda/Dockerfile"
-elif [[ "$image" == *-rocm* ]]; then
-  DOCKERFILE="${OS}-rocm/Dockerfile"
 fi
 
 if [[ "$image" == *-trusty* ]]; then
@@ -178,16 +176,6 @@ case "$image" in
     DB=yes
     VISION=yes
     ;;
-  pytorch-linux-xenial-rocm-py3.6-clang7)
-    ANACONDA_PYTHON_VERSION=3.6
-    CLANG_VERSION=7
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    ROCM=yes
-    # newer cmake version required
-    CMAKE_VERSION=3.6.3
-    ;;
 esac
 
 # Set Jenkins UID and GID if running Jenkins
@@ -225,7 +213,6 @@ docker build \
        --build-arg "CMAKE_VERSION=${CMAKE_VERSION:-}" \
        --build-arg "NINJA_VERSION=${NINJA_VERSION:-}" \
        --build-arg "KATEX=${KATEX:-}" \
-       --build-arg "ROCM=${ROCM:-}" \
        -f $(dirname ${DOCKERFILE})/Dockerfile \
        -t "$tmp_tag" \
        "$@" \

.circleci/verbatim-sources/workflows-ecr-gc.yml

@@ -14,7 +14,7 @@
       - ecr_gc_job:
             name: ecr_gc_job_for_caffe2
             project: caffe2
-            tags_to_keep: "369,348,345,336,325,324,315,306,301,287,283,276,273,266,253,248,238,230,213"
+            tags_to_keep: "348,345,336,325,324,315,306,301,287,283,276,273,266,253,248,238,230,213"
       - ecr_gc_job:
             name: ecr_gc_job_for_translate
             project: translate

aten/src/ATen/native/cuda/CUDAScalar.cu

@@ -4,10 +4,6 @@
 #include <ATen/cuda/CUDAContext.h>
 #include <cuda.h>
 
-#ifdef __HIP_PLATFORM_HCC__
-#include <hip/hip_version.h>
-#endif
-
 namespace at {
 namespace native {
 
@@ -17,12 +13,8 @@ Scalar _local_scalar_dense_cuda(const Tensor& self) {
     at::ScalarType::Half, at::ScalarType::Bool, at::ScalarType::BFloat16, self.scalar_type(), "_local_scalar_dense_cuda", [&] {
         scalar_t value;
         cudaStream_t stream = at::cuda::getCurrentCUDAStream();
-#if HIP_VERSION >= 301
-        AT_CUDA_CHECK(hipMemcpyWithStream(&value, self.data_ptr<scalar_t>(), sizeof(scalar_t), cudaMemcpyDeviceToHost, stream));
-#else
         AT_CUDA_CHECK(cudaMemcpyAsync(&value, self.data_ptr<scalar_t>(), sizeof(scalar_t), cudaMemcpyDeviceToHost, stream));
         AT_CUDA_CHECK(cudaStreamSynchronize(stream));
-#endif
         r = Scalar(value);
       });
   return r;

aten/src/ATen/native/cuda/Copy.cu

@@ -9,10 +9,6 @@
 #include <ATen/native/cuda/Loops.cuh>
 #include <THC/THC.h>
 
-#ifdef __HIP_PLATFORM_HCC__
-#include <hip/hip_version.h>
-#endif
-
 namespace at {
 namespace native {
 
@@ -181,17 +177,13 @@ static void copy_kernel_cuda(TensorIterator& iter, bool non_blocking) {
   int64_t nbytes = iter.numel() * iter.element_size(0);
   CUDAStream stream = getCurrentCUDAStream();
 
+  AT_CUDA_CHECK(cudaMemcpyAsync(dst, src, nbytes, kind, stream));
+
   if (non_blocking) {
-    AT_CUDA_CHECK(cudaMemcpyAsync(dst, src, nbytes, kind, stream));
     void* ptr = (dst_device == kCPU ? dst : src);
     AT_CUDA_CHECK(THCCachingHostAllocator_recordEvent(ptr, stream));
   } else {
-#if HIP_VERSION >= 301
-    AT_CUDA_CHECK(hipMemcpyWithStream(dst, src, nbytes, kind, stream));
-#else
-    AT_CUDA_CHECK(cudaMemcpyAsync(dst, src, nbytes, kind, stream));
     AT_CUDA_CHECK(cudaStreamSynchronize(stream));
-#endif
   }
 }
 

aten/src/THC/THCReduceAll.cuh

@@ -12,10 +12,6 @@
 #include <THC/THCReduceApplyUtils.cuh>
 #include <c10/macros/Macros.h>
 
-#ifdef __HIP_PLATFORM_HCC__
-#include <hip/hip_version.h>
-#endif
-
 // Size per each reduction block
 #define THC_REDUCE_ALL_BLOCK_SIZE 1024L
 
@@ -324,20 +320,12 @@ bool THC_reduceAll(THCState* state,
   // the host (synchronous!)
   if (!outOnDevice) {
     cudaStream_t stream = c10::cuda::getCurrentCUDAStream();
-#if HIP_VERSION >= 301
-    THCudaCheck(hipMemcpyWithStream(out,
-                                    devOut,
-                                    sizeof(AccT),
-                                    cudaMemcpyDeviceToHost,
-                                    stream));
-#else
     THCudaCheck(cudaMemcpyAsync(out,
                                 devOut,
                                 sizeof(AccT),
                                 cudaMemcpyDeviceToHost,
                                 stream));
     THCudaCheck(cudaStreamSynchronize(stream));
-#endif
   }
 
   if (freeDevOut) {

aten/src/THC/generic/THCStorage.cpp

@@ -5,10 +5,6 @@
 #include <c10/util/intrusive_ptr.h>
 #include <c10/util/typeid.h>
 
-#ifdef __HIP_PLATFORM_HCC__
-#include <hip/hip_version.h>
-#endif
-
 scalar_t* THCStorage_(data)(THCState *state, const THCStorage *self)
 {
   return self->data<scalar_t>();
@@ -28,31 +24,20 @@ void THCStorage_(set)(THCState *state, THCStorage *self, ptrdiff_t index, scalar
 {
   THArgCheck((index >= 0) && (index < self->numel()), 2, "index out of bounds");
   cudaStream_t stream = c10::cuda::getCurrentCUDAStream();
-#if HIP_VERSION >= 301
-  THCudaCheck(hipMemcpyWithStream(THCStorage_(data)(state, self) + index, &value, sizeof(scalar_t),
-                                  cudaMemcpyHostToDevice,
-                                  stream));
-#else
   THCudaCheck(cudaMemcpyAsync(THCStorage_(data)(state, self) + index, &value, sizeof(scalar_t),
                               cudaMemcpyHostToDevice,
                               stream));
   THCudaCheck(cudaStreamSynchronize(stream));
-#endif
 }
 
 scalar_t THCStorage_(get)(THCState *state, const THCStorage *self, ptrdiff_t index)
 {
   THArgCheck((index >= 0) && (index < self->numel()), 2, "index out of bounds");
   scalar_t value;
   cudaStream_t stream = c10::cuda::getCurrentCUDAStream();
-#if HIP_VERSION >= 301
-  THCudaCheck(hipMemcpyWithStream(&value, THCStorage_(data)(state, self) + index, sizeof(scalar_t),
-                                  cudaMemcpyDeviceToHost, stream));
-#else
   THCudaCheck(cudaMemcpyAsync(&value, THCStorage_(data)(state, self) + index, sizeof(scalar_t),
                               cudaMemcpyDeviceToHost, stream));
   THCudaCheck(cudaStreamSynchronize(stream));
-#endif
   return value;
 }