ROCm
diff --git a/‎.github/scripts/validate_binaries.sh‎
Lines changed: 22 additions & 2 deletions b/‎.github/scripts/validate_binaries.sh‎
Lines changed: 22 additions & 2 deletions
diff --git a/‎.github/scripts/validate_pipy.sh‎
Lines changed: 14 additions & 6 deletions b/‎.github/scripts/validate_pipy.sh‎
Lines changed: 14 additions & 6 deletions
diff --git a/‎.github/scripts/validate_poetry.sh‎
Lines changed: 15 additions & 11 deletions b/‎.github/scripts/validate_poetry.sh‎
Lines changed: 15 additions & 11 deletions
diff --git a/‎.github/scripts/validate_test_ops.sh‎
Lines changed: 38 additions & 0 deletions b/‎.github/scripts/validate_test_ops.sh‎
Lines changed: 38 additions & 0 deletions
diff --git a/‎.github/workflows/build-conda-images.yml‎
Lines changed: 1 addition & 1 deletion b/‎.github/workflows/build-conda-images.yml‎
Lines changed: 1 addition & 1 deletion
diff --git a/‎.github/workflows/build-libtorch-images.yml‎
Lines changed: 2 additions & 2 deletions b/‎.github/workflows/build-libtorch-images.yml‎
Lines changed: 2 additions & 2 deletions
diff --git a/‎.github/workflows/build-manywheel-images.yml‎
Lines changed: 21 additions & 2 deletions b/‎.github/workflows/build-manywheel-images.yml‎
Lines changed: 21 additions & 2 deletions
diff --git a/‎.github/workflows/lint.yml‎
Lines changed: 35 additions & 0 deletions b/‎.github/workflows/lint.yml‎
Lines changed: 35 additions & 0 deletions
diff --git a/‎.github/workflows/test-check-binary.yml‎
Lines changed: 37 additions & 0 deletions b/‎.github/workflows/test-check-binary.yml‎
Lines changed: 37 additions & 0 deletions
@@ -2,10 +2,25 @@ if [[ ${MATRIX_PACKAGE_TYPE} == "libtorch" ]]; then
     curl ${MATRIX_INSTALLATION} -o libtorch.zip
     unzip libtorch.zip
 else
+
+    # Conda pinned see issue: https://github.com/ContinuumIO/anaconda-issues/issues/13350
+    conda install -y conda=23.11.0
     # Please note ffmpeg is required for torchaudio, see https://github.com/pytorch/pytorch/issues/96159
     conda create -y -n ${ENV_NAME} python=${MATRIX_PYTHON_VERSION} numpy ffmpeg
     conda activate ${ENV_NAME}
     INSTALLATION=${MATRIX_INSTALLATION/"conda install"/"conda install -y"}
+    TEST_SUFFIX=""
+    if [[ ${TORCH_ONLY} == 'true' ]]; then
+        INSTALLATION=${INSTALLATION/"torchvision torchaudio"/""}
+        TEST_SUFFIX=" --package torchonly"
+    fi
+
+    # if RELESE version is passed as parameter - install speific version
+    if [[ ! -z ${RELEASE_VERSION} ]]; then
+          INSTALLATION=${INSTALLATION/"torch "/"torch==${RELEASE_VERSION} "}
+          INSTALLATION=${INSTALLATION/"-y pytorch "/"-y pytorch==${RELEASE_VERSION} "}
+          INSTALLATION=${INSTALLATION/"::pytorch "/"::pytorch==${RELEASE_VERSION} "}
+    fi
 
     export OLD_PATH=${PATH}
     # Workaround macos-arm64 runners. Issue: https://github.com/pytorch/test-infra/issues/4342
@@ -26,15 +41,20 @@ else
     fi
 
     if [[ ${TARGET_OS} == 'windows' ]]; then
-        python  ./test/smoke_test/smoke_test.py
+        python  ./test/smoke_test/smoke_test.py ${TEST_SUFFIX}
     else
-        python3  ./test/smoke_test/smoke_test.py
+        python3  ./test/smoke_test/smoke_test.py ${TEST_SUFFIX}
     fi
 
     if [[ ${TARGET_OS} == 'macos-arm64' ]]; then
         export PATH=${OLD_PATH}
     fi
 
+    # We are only interested in CUDA tests and Python 3.8-3.11. Not all requirement libraries are available for 3.12 yet.
+    if [[ ${INCLUDE_TEST_OPS:-} == 'true' &&  ${MATRIX_GPU_ARCH_TYPE} == 'cuda' && ${MATRIX_PYTHON_VERSION} != "3.12" ]]; then
+        source ./.github/scripts/validate_test_ops.sh
+    fi
+
     conda deactivate
     conda env remove -n ${ENV_NAME}
 fi
@@ -1,12 +1,20 @@
-conda create -yp ${ENV_NAME}_pypi python=${MATRIX_PYTHON_VERSION} numpy ffmpeg
+conda create -yn ${ENV_NAME}_pypi python=${MATRIX_PYTHON_VERSION} numpy ffmpeg
+conda activate ${ENV_NAME}_pypi
 
-if [[ ${MATRIX_CHANNEL} != "release" ]]; then
-    conda run -p ${ENV_NAME}_pypi pip3 install --pre torch --index-url "https://download.pytorch.org/whl/${MATRIX_CHANNEL}/${MATRIX_DESIRED_CUDA}_pypi_cudnn"
-    conda run -p ${ENV_NAME}_pypi pip3 install --pre torchvision torchaudio --index-url "https://download.pytorch.org/whl/${MATRIX_CHANNEL}/${MATRIX_DESIRED_CUDA}"
+TEST_SUFFIX=""
+RELEASE_SUFFIX=""
+# if RELESE version is passed as parameter - install speific version
+if [[ ! -z ${RELEASE_VERSION} ]]; then
+    RELEASE_SUFFIX="==${RELEASE_VERSION}"
+fi
+
+if [[ ${TORCH_ONLY} == 'true' ]]; then
+    TEST_SUFFIX=" --package torchonly"
+    pip3 install torch${RELEASE_SUFFIX}
 else
-    conda run -p ${ENV_NAME}_pypi pip3 install torch torchvision torchaudio
+    pip3 install torch${RELEASE_SUFFIX} torchvision torchaudio
 fi
 
-conda run -p ${ENV_NAME}_pypi python ./test/smoke_test/smoke_test.py
+python ./test/smoke_test/smoke_test.py ${TEST_SUFFIX} --runtime-error-check disabled
 conda deactivate
 conda env remove -p ${ENV_NAME}_pypi
@@ -8,20 +8,24 @@ poetry --version
 poetry new test_poetry
 cd test_poetry
 
-if [[ ${MATRIX_CHANNEL} != "release" ]]; then
-    # Installing poetry from our custom repo. We need to configure it before use and disable authentication
-    export PYTHON_KEYRING_BACKEND=keyring.backends.null.Keyring
-    poetry source add --priority=explicit domains "https://download.pytorch.org/whl/${MATRIX_CHANNEL}/${MATRIX_DESIRED_CUDA}"
-    poetry source add --priority=supplemental pytorch-nightly "https://download.pytorch.org/whl/${MATRIX_CHANNEL}"
-    poetry source add --priority=supplemental pytorch "https://download.pytorch.org/whl/${MATRIX_CHANNEL}/${MATRIX_DESIRED_CUDA}_pypi_cudnn"
-    poetry --quiet add --source pytorch torch
-    poetry --quiet add --source domains torchvision torchaudio
+TEST_SUFFIX=""
+if [[ ${TORCH_ONLY} == 'true' ]]; then
+    TEST_SUFFIX=" --package torchonly"
+fi
+
+RELEASE_SUFFIX=""
+# if RELESE version is passed as parameter - install speific version
+if [[ ! -z ${RELEASE_VERSION} ]]; then
+    RELEASE_SUFFIX="@${RELEASE_VERSION}"
+fi
+
+if [[ ${TORCH_ONLY} == 'true' ]]; then
+    poetry --quiet add torch${RELEASE_SUFFIX}
 else
-    export PYTHON_KEYRING_BACKEND=keyring.backends.null.Keyring
-    poetry --quiet add torch torchaudio torchvision
+    poetry --quiet add torch${RELEASE_SUFFIX} torchaudio torchvision
 fi
 
-python ../test/smoke_test/smoke_test.py
+python ../test/smoke_test/smoke_test.py ${TEST_SUFFIX} --runtime-error-check disabled
 conda deactivate
 conda env remove -p ${ENV_NAME}_poetry
 cd ..
@@ -0,0 +1,38 @@
+#!/bin/bash
+
+set -eux -o pipefail
+
+retry () {
+    $*  || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
+}
+
+BRANCH=""
+if [[ ${MATRIX_CHANNEL} == "test" || ${MATRIX_CHANNEL} == "release" ]]; then
+    SHORT_VERSION=${MATRIX_STABLE_VERSION%.*}
+    BRANCH="--branch release/${SHORT_VERSION}"
+fi
+
+
+# Clone the Pytorch branch
+retry git clone ${BRANCH} --depth 1 https://github.com/pytorch/pytorch.git
+retry git submodule update --init --recursive
+pushd pytorch
+
+pip install expecttest numpy pyyaml jinja2 packaging hypothesis unittest-xml-reporting scipy
+
+# Run pytorch cuda wheels validation
+# Detect ReduceLogicKernel (ReduceOp and kernel) IMA
+python test/test_ops.py -k test_dtypes_all_cuda
+# Detect BinaryMulKernel (elementwise binary functor internal mul) IMA
+python test/test_torch.py -k test_index_reduce_reduce_prod_cuda_int32
+# Detect BinaryBitwiseOpsKernels (at::native::BitwiseAndFunctor) IMA
+python test/test_binary_ufuncs.py -k test_contig_vs_every_other___rand___cuda_int32
+# Detect MaxMinElementwiseKernel (maximum) IMA
+python test/test_schema_check.py -k test_schema_correctness_clamp_cuda_int8
+
+pushd /tmp
+# Detect StepKernel (nextafter) IMA
+python -c "import torch; print(torch.nextafter(torch.tensor([-4.5149, -5.9053, -0.9516, -2.3615,  1.5591], device='cuda:0'), torch.tensor(3.8075, device='cuda:0')))"
+# Detect BinaryGeometricKernels (atan2) IMA
+python -c "import torch; x = (torch.randn((2,1,1), dtype=torch.float, device='cuda')*5).to(torch.float32); y=(torch.randn((), dtype=torch.float, device='cuda')*5).to(torch.float32); print(torch.atan2(x,y))"
+popd
@@ -28,7 +28,7 @@ env:
 
 jobs:
   build-docker:
-    runs-on: ubuntu-22.04
+    runs-on: linux.12xlarge.ephemeral
     strategy:
       matrix:
         cuda_version: ["11.8", "12.1", "cpu"]
 
@@ -30,7 +30,7 @@ env:
 
 jobs:
   build-docker-cuda:
-    runs-on: ubuntu-22.04
+    runs-on: linux.12xlarge.ephemeral
     strategy:
       matrix:
         cuda_version: ["12.1", "11.8"]
@@ -52,7 +52,7 @@ jobs:
     runs-on: linux.12xlarge
     strategy:
       matrix:
-        rocm_version: ["5.5", "5.6"]
+        rocm_version: ["5.7", "6.0"]
     env:
       GPU_ARCH_TYPE: rocm
       GPU_ARCH_VERSION: ${{ matrix.rocm_version }}
 
@@ -12,13 +12,15 @@ on:
     paths:
       - .github/workflows/build-manywheel-images.yml
       - manywheel/Dockerfile
+      - manywheel/Dockerfile_aarch64
       - manywheel/Dockerfile_cxx11-abi
       - manywheel/build_docker.sh
       - 'common/*'
   pull_request:
     paths:
       - .github/workflows/build-manywheel-images.yml
       - manywheel/Dockerfile
+      - manywheel/Dockerfile_aarch64
       - manywheel/Dockerfile_cxx11-abi
       - 'common/*'
       - manywheel/build_docker.sh
@@ -32,14 +34,16 @@ env:
 
 jobs:
   build-docker-cuda:
-    runs-on: ubuntu-22.04
+    runs-on: linux.12xlarge.ephemeral
     strategy:
       matrix:
         cuda_version: ["12.1", "11.8"]
     env:
       GPU_ARCH_TYPE: cuda
       GPU_ARCH_VERSION: ${{ matrix.cuda_version }}
     steps:
+      - name: Purge tools folder (free space for build)
+        run: rm -rf /opt/hostedtoolcache
       - name: Checkout PyTorch builder
         uses: actions/checkout@v3
       - name: Authenticate if WITH_PUSH
@@ -54,7 +58,7 @@ jobs:
     runs-on: linux.12xlarge
     strategy:
       matrix:
-        rocm_version: ["5.5", "5.6"]
+        rocm_version: ["5.7", "6.0"]
     env:
       GPU_ARCH_TYPE: rocm
       GPU_ARCH_VERSION: ${{ matrix.rocm_version }}
@@ -82,6 +86,21 @@ jobs:
       - name: Build Docker Image
         run: |
           manywheel/build_docker.sh
+  build-docker-cpu-aarch64:
+    runs-on: linux.arm64.2xlarge
+    env:
+      GPU_ARCH_TYPE: cpu-aarch64
+    steps:
+      - name: Checkout PyTorch
+        uses: actions/checkout@v3
+      - name: Authenticate if WITH_PUSH
+        run: |
+          if [[ "${WITH_PUSH}" == true ]]; then
+            echo "${DOCKER_TOKEN}" | docker login -u "${DOCKER_ID}" --password-stdin
+          fi
+      - name: Build Docker Image
+        run: |
+          manywheel/build_docker.sh
   build-docker-cpu-cxx11-abi:
     runs-on: ubuntu-22.04
     env:
 
@@ -0,0 +1,35 @@
+name: Lint
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}
+  cancel-in-progress: true
+
+jobs:
+  lintrunner:
+    name: lintrunner
+
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v3
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: "3.11"
+      - name: Install Lintrunner
+        run: |
+          pip install lintrunner
+          lintrunner init
+      - name: Run lintrunner on all files - Linux
+        run: |
+          set +e
+          if ! lintrunner -v --force-color --all-files --tee-json=lint.json; then
+              echo ""
+              echo -e "\e[1m\e[36mYou can reproduce these results locally by using \`lintrunner -m main\`.\e[0m"
+              exit 1
+          fi
@@ -0,0 +1,37 @@
+name: Test check_binary
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    paths:
+      - .github/workflows/test-check-binary.yml
+      - check_binary.sh
+      - test/smoke_test/smoke_test.py
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}
+  cancel-in-progress: true
+
+jobs:
+  check_binary_linux_cpu:
+    name: Test check_binary.sh for Linux CPU
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      repository: "pytorch/builder"
+      docker-image: python:3.11
+      script: |
+          pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu
+          DESIRED_PYTHON=3.11 DESIRED_CUDA=cpu PACKAGE_TYPE=manywheel ./check_binary.sh
+
+  check_binary_linux_cuda:
+    name: Test check_binary.sh for Linux CUDA
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    with:
+      repository: "pytorch/builder"
+      runner: linux.4xlarge.nvidia.gpu
+      docker-image: python:3.11
+      script: |
+          pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu121
+          DESIRED_PYTHON=3.11 DESIRED_CUDA=cu121 PACKAGE_TYPE=manywheel ./check_binary.sh