Triton wheel build using 2.3.x branch (#122403)

* Triton build 2.3.x

* Revert "[Release Only] Build triton using pinned version rather branch (#121765)"

This reverts commit d69c4219127e2cf5d9637b0daacc0a24e65f8133.

* Triton wheel change

* release

.bazelignore

@@ -1,4 +1,3 @@
 # We do not use this library in our Bazel build. It contains an
 # infinitely recursing symlink that makes Bazel very unhappy.
 third_party/ittapi/
-third_party/opentelemetry-cpp

.ci/docker/aotriton_version.txt

@@ -1,5 +0,0 @@
-0.6b
-manylinux_2_17
-rocm6.1
-7f07e8a1cb1f99627eb6d77f5c0e9295c775f3c7
-77c29fa3f3b614e187d7213d745e989a92708cee2bc6020419ab49019af399d1

.ci/docker/build.sh

@@ -84,30 +84,16 @@ fi
 # CMake 3.18 is needed to support CUDA17 language variant
 CMAKE_VERSION=3.18.5
 
-_UCX_COMMIT=7bb2722ff2187a0cad557ae4a6afa090569f83fb
-_UCC_COMMIT=20eae37090a4ce1b32bcce6144ccad0b49943e0b
+_UCX_COMMIT=00bcc6bb18fc282eb160623b4c0d300147f579af
+_UCC_COMMIT=7cb07a76ccedad7e56ceb136b865eb9319c258ea
 
 # It's annoying to rename jobs every time you want to rewrite a
 # configuration, so we hardcode everything here rather than do it
 # from scratch
 case "$image" in
-  pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9)
-    CUDA_VERSION=12.4.0
-    CUDNN_VERSION=9
-    ANACONDA_PYTHON_VERSION=3.10
-    GCC_VERSION=9
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    KATEX=yes
-    UCX_COMMIT=${_UCX_COMMIT}
-    UCC_COMMIT=${_UCC_COMMIT}
-    CONDA_CMAKE=yes
-    TRITON=yes
-    ;;
-  pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9)
+  pytorch-linux-focal-cuda12.1-cudnn8-py3-gcc9)
     CUDA_VERSION=12.1.1
-    CUDNN_VERSION=9
+    CUDNN_VERSION=8
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
     PROTOBUF=yes
@@ -119,24 +105,9 @@ case "$image" in
     CONDA_CMAKE=yes
     TRITON=yes
     ;;
-  pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.4.0
-    CUDNN_VERSION=9
-    ANACONDA_PYTHON_VERSION=3.10
-    GCC_VERSION=9
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    KATEX=yes
-    UCX_COMMIT=${_UCX_COMMIT}
-    UCC_COMMIT=${_UCC_COMMIT}
-    CONDA_CMAKE=yes
-    TRITON=yes
-    INDUCTOR_BENCHMARKS=yes
-    ;;
-  pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks)
+  pytorch-linux-focal-cuda12.1-cudnn8-py3-gcc9-inductor-benchmarks)
     CUDA_VERSION=12.1.1
-    CUDNN_VERSION=9
+    CUDNN_VERSION=8
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
     PROTOBUF=yes
@@ -149,39 +120,9 @@ case "$image" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-focal-cuda12.1-cudnn9-py3.12-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.1.1
-    CUDNN_VERSION=9
-    ANACONDA_PYTHON_VERSION=3.12
-    GCC_VERSION=9
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    KATEX=yes
-    UCX_COMMIT=${_UCX_COMMIT}
-    UCC_COMMIT=${_UCC_COMMIT}
-    CONDA_CMAKE=yes
-    TRITON=yes
-    INDUCTOR_BENCHMARKS=yes
-    ;;
-  pytorch-linux-focal-cuda12.4-cudnn9-py3.12-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.4.0
-    CUDNN_VERSION=9
-    ANACONDA_PYTHON_VERSION=3.12
-    GCC_VERSION=9
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    KATEX=yes
-    UCX_COMMIT=${_UCX_COMMIT}
-    UCC_COMMIT=${_UCC_COMMIT}
-    CONDA_CMAKE=yes
-    TRITON=yes
-    INDUCTOR_BENCHMARKS=yes
-    ;;
-  pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9)
+  pytorch-linux-focal-cuda11.8-cudnn8-py3-gcc9)
     CUDA_VERSION=11.8.0
-    CUDNN_VERSION=9
+    CUDNN_VERSION=8
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
     PROTOBUF=yes
@@ -193,37 +134,9 @@ case "$image" in
     CONDA_CMAKE=yes
     TRITON=yes
     ;;
-  pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9)
-    CUDA_VERSION=12.4.0
-    CUDNN_VERSION=9
-    ANACONDA_PYTHON_VERSION=3.10
-    GCC_VERSION=9
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    KATEX=yes
-    UCX_COMMIT=${_UCX_COMMIT}
-    UCC_COMMIT=${_UCC_COMMIT}
-    CONDA_CMAKE=yes
-    TRITON=yes
-    ;;
-  pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9)
+  pytorch-linux-focal-cuda12.1-cudnn8-py3-gcc9)
     CUDA_VERSION=12.1.1
-    CUDNN_VERSION=9
-    ANACONDA_PYTHON_VERSION=3.10
-    GCC_VERSION=9
-    PROTOBUF=yes
-    DB=yes
-    VISION=yes
-    KATEX=yes
-    UCX_COMMIT=${_UCX_COMMIT}
-    UCC_COMMIT=${_UCC_COMMIT}
-    CONDA_CMAKE=yes
-    TRITON=yes
-    ;;
-  pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9)
-    CUDA_VERSION=12.4.0
-    CUDNN_VERSION=9
+    CUDNN_VERSION=8
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
     PROTOBUF=yes
@@ -291,7 +204,7 @@ case "$image" in
     PROTOBUF=yes
     DB=yes
     VISION=yes
-    ROCM_VERSION=6.0
+    ROCM_VERSION=5.7
     NINJA_VERSION=1.9.0
     CONDA_CMAKE=yes
     TRITON=yes
@@ -302,7 +215,7 @@ case "$image" in
     PROTOBUF=yes
     DB=yes
     VISION=yes
-    ROCM_VERSION=6.1
+    ROCM_VERSION=6.0
     NINJA_VERSION=1.9.0
     CONDA_CMAKE=yes
     TRITON=yes
@@ -313,10 +226,9 @@ case "$image" in
     PROTOBUF=yes
     DB=yes
     VISION=yes
-    XPU_VERSION=0.5
+    BASEKIT_VERSION=2024.0.0-49522
     NINJA_VERSION=1.9.0
     CONDA_CMAKE=yes
-    TRITON=yes
     ;;
     pytorch-linux-jammy-py3.8-gcc11-inductor-benchmarks)
     ANACONDA_PYTHON_VERSION=3.8
@@ -330,10 +242,10 @@ case "$image" in
     DOCS=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda11.8-cudnn9-py3.8-clang12)
+  pytorch-linux-jammy-cuda11.8-cudnn8-py3.8-clang12)
     ANACONDA_PYTHON_VERSION=3.8
     CUDA_VERSION=11.8
-    CUDNN_VERSION=9
+    CUDNN_VERSION=8
     CLANG_VERSION=12
     PROTOBUF=yes
     DB=yes
@@ -373,13 +285,6 @@ case "$image" in
     CONDA_CMAKE=yes
     EXECUTORCH=yes
     ;;
-  pytorch-linux-jammy-py3.12-halide)
-    CUDA_VERSION=12.4
-    ANACONDA_PYTHON_VERSION=3.12
-    GCC_VERSION=11
-    CONDA_CMAKE=yes
-    HALIDE=yes
-    ;;
   pytorch-linux-focal-linter)
     # TODO: Use 3.9 here because of this issue https://github.com/python/mypy/issues/13627.
     # We will need to update mypy version eventually, but that's for another day. The task
@@ -387,7 +292,7 @@ case "$image" in
     ANACONDA_PYTHON_VERSION=3.9
     CONDA_CMAKE=yes
     ;;
-  pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-linter)
+  pytorch-linux-jammy-cuda11.8-cudnn8-py3.9-linter)
     ANACONDA_PYTHON_VERSION=3.9
     CUDA_VERSION=11.8
     CONDA_CMAKE=yes
@@ -400,12 +305,6 @@ case "$image" in
     DB=yes
     VISION=yes
     CONDA_CMAKE=yes
-    # snadampal: skipping sccache due to the following issue
-    # https://github.com/pytorch/pytorch/issues/121559
-    SKIP_SCCACHE_INSTALL=yes
-    # snadampal: skipping llvm src build install because the current version
-    # from pytorch/llvm:9.0.1 is x86 specific
-    SKIP_LLVM_SRC_BUILD_INSTALL=yes
     ;;
   *)
     # Catch-all for builds that are not hardcoded.
@@ -454,13 +353,13 @@ tmp_tag=$(basename "$(mktemp -u)" | tr '[:upper:]' '[:lower:]')
 #when using cudnn version 8 install it separately from cuda
 if [[ "$image" == *cuda*  && ${OS} == "ubuntu" ]]; then
   IMAGE_NAME="nvidia/cuda:${CUDA_VERSION}-cudnn${CUDNN_VERSION}-devel-ubuntu${UBUNTU_VERSION}"
-  if [[ ${CUDNN_VERSION} == 9 ]]; then
+  if [[ ${CUDNN_VERSION} == 8 ]]; then
     IMAGE_NAME="nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}"
   fi
 fi
 
 # Build image
-docker build \
+DOCKER_BUILDKIT=1 docker build \
        --no-cache \
        --progress=plain \
        --build-arg "BUILD_ENVIRONMENT=${image}" \
@@ -497,17 +396,14 @@ docker build \
        --build-arg "DOCS=${DOCS}" \
        --build-arg "INDUCTOR_BENCHMARKS=${INDUCTOR_BENCHMARKS}" \
        --build-arg "EXECUTORCH=${EXECUTORCH}" \
-       --build-arg "HALIDE=${HALIDE}" \
-       --build-arg "XPU_VERSION=${XPU_VERSION}" \
+       --build-arg "BASEKIT_VERSION=${BASEKIT_VERSION}" \
        --build-arg "ACL=${ACL:-}" \
-       --build-arg "SKIP_SCCACHE_INSTALL=${SKIP_SCCACHE_INSTALL:-}" \
-       --build-arg "SKIP_LLVM_SRC_BUILD_INSTALL=${SKIP_LLVM_SRC_BUILD_INSTALL:-}" \
        -f $(dirname ${DOCKERFILE})/Dockerfile \
        -t "$tmp_tag" \
        "$@" \
        .
 
-# NVIDIA dockers for RC releases use tag names like `11.0-cudnn9-devel-ubuntu18.04-rc`,
+# NVIDIA dockers for RC releases use tag names like `11.0-cudnn8-devel-ubuntu18.04-rc`,
 # for this case we will set UBUNTU_VERSION to `18.04-rc` so that the Dockerfile could
 # find the correct image. As a result, here we have to replace the
 #   "$UBUNTU_VERSION" == "18.04-rc"

.ci/docker/centos-rocm/Dockerfile

@@ -62,7 +62,7 @@ RUN if [ -n "${DB}" ]; then bash ./install_db.sh; fi
 RUN rm install_db.sh
 ENV INSTALLED_DB ${DB}
 
-# (optional) Install vision packages like OpenCV
+# (optional) Install vision packages like OpenCV and ffmpeg
 ARG VISION
 COPY ./common/install_vision.sh ./common/cache_vision_models.sh ./common/common_utils.sh ./
 RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
@@ -77,9 +77,6 @@ RUN rm install_rocm.sh
 COPY ./common/install_rocm_magma.sh install_rocm_magma.sh
 RUN bash ./install_rocm_magma.sh
 RUN rm install_rocm_magma.sh
-COPY ./common/install_amdsmi.sh install_amdsmi.sh
-RUN bash ./install_amdsmi.sh
-RUN rm install_amdsmi.sh
 ENV PATH /opt/rocm/bin:$PATH
 ENV PATH /opt/rocm/hcc/bin:$PATH
 ENV PATH /opt/rocm/hip/bin:$PATH
@@ -113,13 +110,6 @@ COPY triton_version.txt triton_version.txt
 RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi
 RUN rm install_triton.sh common_utils.sh triton-rocm.txt triton_version.txt
 
-# Install AOTriton (Early fail)
-COPY ./aotriton_version.txt aotriton_version.txt
-COPY ./common/common_utils.sh common_utils.sh
-COPY ./common/install_aotriton.sh install_aotriton.sh
-RUN ["/bin/bash", "-c", "./install_aotriton.sh /opt/rocm && rm -rf install_aotriton.sh aotriton_version.txt common_utils.sh"]
-ENV AOTRITON_INSTALLED_PREFIX /opt/rocm/aotriton
-
 # Install ccache/sccache (do this last, so we get priority in PATH)
 COPY ./common/install_cache.sh install_cache.sh
 ENV PATH /opt/cache/bin:$PATH

.ci/docker/ci_commit_pins/executorch.txt

@@ -1 +1 @@
-9d859653ae916d0a72f6b2b5c5925bed38832140
+e2a8f9548aecb62a68e264607174a7d207ed2929

.ci/docker/ci_commit_pins/halide.txt

@@ -1 +0,0 @@
-340136fec6d3ebc73e7a19eba1663e9b0ba8ab2d

.ci/docker/ci_commit_pins/triton-rocm.txt

@@ -1 +1 @@
-21eae954efa5bf584da70324b640288c3ee7aede
+0a22a91d04c2b4a029a69a198eac390089c3e891

.ci/docker/ci_commit_pins/triton-xpu.txt

@@ -1 +0,0 @@
-1b2f15840e0d70eec50d84c7a0575cb835524def

.ci/docker/ci_commit_pins/triton.txt

@@ -1 +1 @@
-dedb7bdf339a3546896d4820366ca562c586bfa0
+79c6c9b209a5692b9a895398f4f3a033f8f80415

.ci/docker/common/install_acl.sh

@@ -1,6 +1,6 @@
 set -euo pipefail
 
-readonly version=v24.04
+readonly version=v23.08
 readonly src_host=https://review.mlplatform.org/ml
 readonly src_repo=ComputeLibrary
 

.ci/docker/common/install_amdsmi.sh

@@ -1,5 +0,0 @@
-#!/bin/bash
-
-set -ex
-
-cd /opt/rocm/share/amd_smi && pip install .

.ci/docker/common/install_aotriton.sh

@@ -1,23 +0,0 @@
-#!/bin/bash
-
-set -ex
-
-source "$(dirname "${BASH_SOURCE[0]}")/common_utils.sh"
-
-TARBALL='aotriton.tar.bz2'
-# This read command alwasy returns with exit code 1
-read -d "\n" VER MANYLINUX ROCMBASE PINNED_COMMIT SHA256 < aotriton_version.txt || true
-ARCH=$(uname -m)
-AOTRITON_INSTALL_PREFIX="$1"
-AOTRITON_URL="https://github.com/ROCm/aotriton/releases/download/${VER}/aotriton-${VER}-${MANYLINUX}_${ARCH}-${ROCMBASE}-shared.tar.bz2"
-
-cd "${AOTRITON_INSTALL_PREFIX}"
-# Must use -L to follow redirects
-curl -L --retry 3 -o "${TARBALL}" "${AOTRITON_URL}"
-ACTUAL_SHA256=$(sha256sum "${TARBALL}" | cut -d " " -f 1)
-if [ "${SHA256}" != "${ACTUAL_SHA256}" ]; then
-  echo -n "Error: The SHA256 of downloaded tarball is ${ACTUAL_SHA256},"
-  echo " which does not match the expected value ${SHA256}."
-  exit
-fi
-tar xf "${TARBALL}" && rm -rf "${TARBALL}"

.ci/docker/common/install_base.sh

@@ -3,7 +3,7 @@
 set -ex
 
 install_ubuntu() {
-  # NVIDIA dockers for RC releases use tag names like `11.0-cudnn9-devel-ubuntu18.04-rc`,
+  # NVIDIA dockers for RC releases use tag names like `11.0-cudnn8-devel-ubuntu18.04-rc`,
   # for this case we will set UBUNTU_VERSION to `18.04-rc` so that the Dockerfile could
   # find the correct image. As a result, here we have to check for
   #   "$UBUNTU_VERSION" == "18.04"*
@@ -113,6 +113,7 @@ install_centos() {
     glibc-devel \
     glibc-headers \
     glog-devel \
+    hiredis-devel \
     libstdc++-devel \
     libsndfile-devel \
     make \

.ci/docker/common/install_conda.sh

@@ -85,7 +85,7 @@ fi
   else
     CONDA_COMMON_DEPS="astunparse pyyaml mkl=2021.4.0 mkl-include=2021.4.0 setuptools"
 
-    if [ "$ANACONDA_PYTHON_VERSION" = "3.11" ] || [ "$ANACONDA_PYTHON_VERSION" = "3.12" ] || [ "$ANACONDA_PYTHON_VERSION" = "3.13" ]; then
+    if [ "$ANACONDA_PYTHON_VERSION" = "3.11" ] || [ "$ANACONDA_PYTHON_VERSION" = "3.12" ]; then
       conda_install numpy=1.26.0 ${CONDA_COMMON_DEPS}
     else
       conda_install numpy=1.21.2 ${CONDA_COMMON_DEPS}

.ci/docker/common/install_cudnn.sh

@@ -1,18 +1,20 @@
 #!/bin/bash
 
-if [[ -n "${CUDNN_VERSION}" ]]; then
+if [[ ${CUDNN_VERSION} == 8 ]]; then
     # cuDNN license: https://developer.nvidia.com/cudnn/license_agreement
     mkdir tmp_cudnn
     pushd tmp_cudnn
-    if [[ ${CUDA_VERSION:0:2} == "12" ]]; then
-        CUDNN_NAME="cudnn-linux-x86_64-9.1.0.70_cuda12-archive"
-    elif [[ ${CUDA_VERSION:0:2} == "11" ]]; then
-        CUDNN_NAME="cudnn-linux-x86_64-9.1.0.70_cuda11-archive"
+    if [[ ${CUDA_VERSION:0:4} == "12.1" ]]; then
+        CUDNN_NAME="cudnn-linux-x86_64-8.9.2.26_cuda12-archive"
+        curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/linux-x86_64/${CUDNN_NAME}.tar.xz
+    elif [[ ${CUDA_VERSION:0:4} == "11.8" ]]; then
+        CUDNN_NAME="cudnn-linux-x86_64-8.7.0.84_cuda11-archive"
+        curl --retry 3 -OLs https://developer.download.nvidia.com/compute/redist/cudnn/v8.7.0/local_installers/11.8/${CUDNN_NAME}.tar.xz
     else
         print "Unsupported CUDA version ${CUDA_VERSION}"
         exit 1
     fi
-    curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/linux-x86_64/${CUDNN_NAME}.tar.xz
+
     tar xf ${CUDNN_NAME}.tar.xz
     cp -a ${CUDNN_NAME}/include/* /usr/local/cuda/include/
     cp -a ${CUDNN_NAME}/lib/* /usr/local/cuda/lib64/

.ci/docker/common/install_cusparselt.sh

@@ -5,14 +5,9 @@ set -ex
 # cuSPARSELt license: https://docs.nvidia.com/cuda/cusparselt/license.html
 mkdir tmp_cusparselt && cd tmp_cusparselt
 
-if [[ ${CUDA_VERSION:0:4} =~ ^12\.[1-4]$ ]]; then
-    arch_path='sbsa'
-    export TARGETARCH=${TARGETARCH:-$(uname -m)}
-    if [ ${TARGETARCH} = 'amd64' ] || [ "${TARGETARCH}" = 'x86_64' ]; then
-        arch_path='x86_64'
-    fi
-    CUSPARSELT_NAME="libcusparse_lt-linux-${arch_path}-0.5.2.1-archive"
-    curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-${arch_path}/${CUSPARSELT_NAME}.tar.xz
+if [[ ${CUDA_VERSION:0:4} == "12.1" ]]; then
+    CUSPARSELT_NAME="libcusparse_lt-linux-x86_64-0.5.2.1-archive"
+    curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-x86_64/${CUSPARSELT_NAME}.tar.xz
 elif [[ ${CUDA_VERSION:0:4} == "11.8" ]]; then
     CUSPARSELT_NAME="libcusparse_lt-linux-x86_64-0.4.0.7-archive"
     curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-x86_64/${CUSPARSELT_NAME}.tar.xz

.ci/docker/common/install_db.sh

@@ -4,6 +4,11 @@ set -ex
 
 install_ubuntu() {
   apt-get update
+  apt-get install -y --no-install-recommends \
+          libhiredis-dev \
+          libleveldb-dev \
+          liblmdb-dev \
+          libsnappy-dev
 
   # Cleanup
   apt-get autoclean && apt-get clean
@@ -15,6 +20,12 @@ install_centos() {
   # See http://fedoraproject.org/wiki/EPEL
   yum --enablerepo=extras install -y epel-release
 
+  yum install -y \
+      hiredis-devel \
+      leveldb-devel \
+      lmdb-devel \
+      snappy-devel
+
   # Cleanup
   yum clean all
   rm -rf /var/cache/yum

.ci/docker/common/install_executorch.sh

@@ -37,9 +37,6 @@ install_conda_dependencies() {
 
 install_pip_dependencies() {
   pushd executorch/.ci/docker
-  # Install PyTorch CPU build beforehand to avoid installing the much bigger CUDA
-  # binaries later, ExecuTorch only needs CPU
-  pip_install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
   # Install all Python dependencies
   pip_install -r requirements-ci.txt
   popd
@@ -47,14 +44,13 @@ install_pip_dependencies() {
 
 setup_executorch() {
   pushd executorch
-  # Setup swiftshader and Vulkan SDK which are required to build the Vulkan delegate
-  as_jenkins bash .ci/scripts/setup-vulkan-linux-deps.sh
+  source .ci/scripts/utils.sh
 
-  export PYTHON_EXECUTABLE=python
-  export EXECUTORCH_BUILD_PYBIND=ON
-  export CMAKE_ARGS="-DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON"
+  install_flatc_from_source
+  pip_install .
 
-  as_jenkins .ci/scripts/setup-linux.sh cmake
+  # Make sure that all the newly generate files are owned by Jenkins
+  chown -R jenkins .
   popd
 }
 

.ci/docker/common/install_halide.sh

@@ -1,46 +0,0 @@
-#!/bin/bash
-set -ex
-
-source "$(dirname "${BASH_SOURCE[0]}")/common_utils.sh"
-
-COMMIT=$(get_pinned_commit halide)
-test -n "$COMMIT"
-
-# activate conda to populate CONDA_PREFIX
-test -n "$ANACONDA_PYTHON_VERSION"
-eval "$(conda shell.bash hook)"
-conda activate py_$ANACONDA_PYTHON_VERSION
-
-if [ -n "${UBUNTU_VERSION}" ];then
-    apt update
-    apt-get install -y lld liblld-15-dev libpng-dev libjpeg-dev libgl-dev \
-                  libopenblas-dev libeigen3-dev libatlas-base-dev libzstd-dev
-fi
-
-conda_install numpy scipy imageio cmake ninja
-
-git clone --depth 1 --branch release/16.x --recursive https://github.com/llvm/llvm-project.git
-cmake -DCMAKE_BUILD_TYPE=Release \
-        -DLLVM_ENABLE_PROJECTS="clang" \
-        -DLLVM_TARGETS_TO_BUILD="X86;NVPTX" \
-        -DLLVM_ENABLE_TERMINFO=OFF -DLLVM_ENABLE_ASSERTIONS=ON \
-        -DLLVM_ENABLE_EH=ON -DLLVM_ENABLE_RTTI=ON -DLLVM_BUILD_32_BITS=OFF \
-        -S llvm-project/llvm -B llvm-build -G Ninja
-cmake --build llvm-build
-cmake --install llvm-build --prefix llvm-install
-export LLVM_ROOT=`pwd`/llvm-install
-export LLVM_CONFIG=$LLVM_ROOT/bin/llvm-config
-
-git clone https://github.com/halide/Halide.git
-pushd Halide
-git checkout ${COMMIT} && git submodule update --init --recursive
-pip_install -r requirements.txt
-cmake -G Ninja -DCMAKE_BUILD_TYPE=Release -S . -B build
-cmake --build build
-test -e ${CONDA_PREFIX}/lib/python3 || ln -s python${ANACONDA_PYTHON_VERSION} ${CONDA_PREFIX}/lib/python3
-cmake --install build --prefix ${CONDA_PREFIX}
-chown -R jenkins ${CONDA_PREFIX}
-popd
-rm -rf Halide llvm-build llvm-project llvm-install
-
-python -c "import halide"  # check for errors

.ci/docker/common/install_onnx.sh

@@ -30,17 +30,15 @@ pip_install \
 
 pip_install coloredlogs packaging
 
-pip_install onnxruntime==1.18
-pip_install onnx==1.16.0
+pip_install onnxruntime==1.17.0
+pip_install onnx==1.15.0
 # pip_install "onnxscript@git+https://github.com/microsoft/onnxscript@3e869ef8ccf19b5ebd21c10d3e9c267c9a9fa729" --no-deps
-pip_install onnxscript==0.1.0.dev20240613 --no-deps
-# required by onnxscript
-pip_install ml_dtypes
+pip_install onnxscript==0.1.0.dev20240301 --no-deps
 
 # Cache the transformers model to be used later by ONNX tests. We need to run the transformers
 # package to download the model. By default, the model is cached at ~/.cache/huggingface/hub/
 IMPORT_SCRIPT_FILENAME="/tmp/onnx_import_script.py"
-as_jenkins echo 'import transformers; transformers.AutoModel.from_pretrained("sshleifer/tiny-gpt2"); transformers.AutoTokenizer.from_pretrained("sshleifer/tiny-gpt2"); transformers.AutoModelForSpeechSeq2Seq.from_pretrained("openai/whisper-large-v3");' > "${IMPORT_SCRIPT_FILENAME}"
+as_jenkins echo 'import transformers; transformers.AutoModel.from_pretrained("sshleifer/tiny-gpt2"); transformers.AutoTokenizer.from_pretrained("sshleifer/tiny-gpt2");' > "${IMPORT_SCRIPT_FILENAME}"
 
 # Need a PyTorch version for transformers to work
 pip_install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu

.ci/docker/common/install_protobuf.sh

@@ -11,8 +11,7 @@ mkdir -p $pb_dir
 ln -s /usr/lib64 "$pb_dir/lib64"
 
 curl -LO "https://github.com/protocolbuffers/protobuf/releases/download/v3.17.3/protobuf-all-3.17.3.tar.gz" --retry 3
-
-tar -xvz --no-same-owner -C "$pb_dir" --strip-components 1 -f protobuf-all-3.17.3.tar.gz
+tar -xvz -C "$pb_dir" --strip-components 1 -f protobuf-all-3.17.3.tar.gz
 NPROC=$[$(nproc) - 2]
 pushd "$pb_dir" && ./configure && make -j${NPROC} && make -j${NPROC} check && sudo make -j${NRPOC} install && sudo ldconfig
 popd

.ci/docker/common/install_rocm.sh

@@ -6,6 +6,9 @@ ver() {
     printf "%3d%03d%03d%03d" $(echo "$1" | tr '.' ' ');
 }
 
+# Map ROCm version to AMDGPU version
+declare -A AMDGPU_VERSIONS=( ["5.0"]="21.50" ["5.1.1"]="22.10.1" ["5.2"]="22.20" )
+
 install_ubuntu() {
     apt-get update
     if [[ $UBUNTU_VERSION == 18.04 ]]; then
@@ -23,14 +26,31 @@ install_ubuntu() {
     apt-get install -y libc++1
     apt-get install -y libc++abi1
 
-    # Add amdgpu repository
-    UBUNTU_VERSION_NAME=`cat /etc/os-release | grep UBUNTU_CODENAME | awk -F= '{print $2}'`
-    echo "deb [arch=amd64] https://repo.radeon.com/amdgpu/${ROCM_VERSION}/ubuntu ${UBUNTU_VERSION_NAME} main" > /etc/apt/sources.list.d/amdgpu.list
+    if [[ $(ver $ROCM_VERSION) -ge $(ver 4.5) ]]; then
+        # Add amdgpu repository
+        UBUNTU_VERSION_NAME=`cat /etc/os-release | grep UBUNTU_CODENAME | awk -F= '{print $2}'`
+        local amdgpu_baseurl
+        if [[ $(ver $ROCM_VERSION) -ge $(ver 5.3) ]]; then
+          amdgpu_baseurl="https://repo.radeon.com/amdgpu/${ROCM_VERSION}/ubuntu"
+        else
+          amdgpu_baseurl="https://repo.radeon.com/amdgpu/${AMDGPU_VERSIONS[$ROCM_VERSION]}/ubuntu"
+        fi
+        echo "deb [arch=amd64] ${amdgpu_baseurl} ${UBUNTU_VERSION_NAME} main" > /etc/apt/sources.list.d/amdgpu.list
+    fi
+
+    ROCM_REPO="ubuntu"
+    if [[ $(ver $ROCM_VERSION) -lt $(ver 4.2) ]]; then
+        ROCM_REPO="xenial"
+    fi
+
+    if [[ $(ver $ROCM_VERSION) -ge $(ver 5.3) ]]; then
+        ROCM_REPO="${UBUNTU_VERSION_NAME}"
+    fi
 
     # Add rocm repository
     wget -qO - http://repo.radeon.com/rocm/rocm.gpg.key | apt-key add -
     local rocm_baseurl="http://repo.radeon.com/rocm/apt/${ROCM_VERSION}"
-    echo "deb [arch=amd64] ${rocm_baseurl} ${UBUNTU_VERSION_NAME} main" > /etc/apt/sources.list.d/rocm.list
+    echo "deb [arch=amd64] ${rocm_baseurl} ${ROCM_REPO} main" > /etc/apt/sources.list.d/rocm.list
     apt-get update --allow-insecure-repositories
 
     DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \
@@ -39,28 +59,34 @@ install_ubuntu() {
                    rocm-libs \
                    rccl \
                    rocprofiler-dev \
-                   roctracer-dev \
-                   amd-smi-lib
-
-    if [[ $(ver $ROCM_VERSION) -ge $(ver 6.1) ]]; then
-        DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated rocm-llvm-dev
-    fi
+                   roctracer-dev
 
     # precompiled miopen kernels added in ROCm 3.5, renamed in ROCm 5.5
     # search for all unversioned packages
     # if search fails it will abort this script; use true to avoid case where search fails
-    MIOPENHIPGFX=$(apt-cache search --names-only miopen-hip-gfx | awk '{print $1}' | grep -F -v . || true)
-    if [[ "x${MIOPENHIPGFX}" = x ]]; then
-      echo "miopen-hip-gfx package not available" && exit 1
+    if [[ $(ver $ROCM_VERSION) -ge $(ver 5.5) ]]; then
+        MIOPENHIPGFX=$(apt-cache search --names-only miopen-hip-gfx | awk '{print $1}' | grep -F -v . || true)
+        if [[ "x${MIOPENHIPGFX}" = x ]]; then
+          echo "miopen-hip-gfx package not available" && exit 1
+        else
+          DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ${MIOPENHIPGFX}
+        fi
     else
-      DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ${MIOPENHIPGFX}
+        MIOPENKERNELS=$(apt-cache search --names-only miopenkernels | awk '{print $1}' | grep -F -v . || true)
+        if [[ "x${MIOPENKERNELS}" = x ]]; then
+          echo "miopenkernels package not available" && exit 1
+        else
+          DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ${MIOPENKERNELS}
+        fi
     fi
 
     # ROCm 6.0 had a regression where journal_mode was enabled on the kdb files resulting in permission errors at runtime
-    for kdb in /opt/rocm/share/miopen/db/*.kdb
-    do
-        sqlite3 $kdb "PRAGMA journal_mode=off; PRAGMA VACUUM;"
-    done
+    if [[ $(ver $ROCM_VERSION) -ge $(ver 6.0) ]]; then
+        for kdb in /opt/rocm/share/miopen/db/*.kdb
+        do
+            sqlite3 $kdb "PRAGMA journal_mode=off; PRAGMA VACUUM;"
+        done
+    fi
 
     # Cleanup
     apt-get autoclean && apt-get clean
@@ -77,19 +103,25 @@ install_centos() {
   yum install -y epel-release
   yum install -y dkms kernel-headers-`uname -r` kernel-devel-`uname -r`
 
-  # Add amdgpu repository
-  local amdgpu_baseurl
-  if [[ $OS_VERSION == 9 ]]; then
-      amdgpu_baseurl="https://repo.radeon.com/amdgpu/${ROCM_VERSION}/rhel/9.0/main/x86_64"
-  else
-      amdgpu_baseurl="https://repo.radeon.com/amdgpu/${ROCM_VERSION}/rhel/7.9/main/x86_64"
+  if [[ $(ver $ROCM_VERSION) -ge $(ver 4.5) ]]; then
+      # Add amdgpu repository
+      local amdgpu_baseurl
+      if [[ $OS_VERSION == 9 ]]; then
+          amdgpu_baseurl="https://repo.radeon.com/amdgpu/${AMDGPU_VERSIONS[$ROCM_VERSION]}/rhel/9.0/main/x86_64"
+      else
+        if [[ $(ver $ROCM_VERSION) -ge $(ver 5.3) ]]; then
+          amdgpu_baseurl="https://repo.radeon.com/amdgpu/${ROCM_VERSION}/rhel/7.9/main/x86_64"
+        else
+          amdgpu_baseurl="https://repo.radeon.com/amdgpu/${AMDGPU_VERSIONS[$ROCM_VERSION]}/rhel/7.9/main/x86_64"
+        fi
+      fi
+      echo "[AMDGPU]" > /etc/yum.repos.d/amdgpu.repo
+      echo "name=AMDGPU" >> /etc/yum.repos.d/amdgpu.repo
+      echo "baseurl=${amdgpu_baseurl}" >> /etc/yum.repos.d/amdgpu.repo
+      echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo
+      echo "gpgcheck=1" >> /etc/yum.repos.d/amdgpu.repo
+      echo "gpgkey=http://repo.radeon.com/rocm/rocm.gpg.key" >> /etc/yum.repos.d/amdgpu.repo
   fi
-  echo "[AMDGPU]" > /etc/yum.repos.d/amdgpu.repo
-  echo "name=AMDGPU" >> /etc/yum.repos.d/amdgpu.repo
-  echo "baseurl=${amdgpu_baseurl}" >> /etc/yum.repos.d/amdgpu.repo
-  echo "enabled=1" >> /etc/yum.repos.d/amdgpu.repo
-  echo "gpgcheck=1" >> /etc/yum.repos.d/amdgpu.repo
-  echo "gpgkey=http://repo.radeon.com/rocm/rocm.gpg.key" >> /etc/yum.repos.d/amdgpu.repo
 
   local rocm_baseurl="http://repo.radeon.com/rocm/yum/${ROCM_VERSION}"
   echo "[ROCm]" > /etc/yum.repos.d/rocm.repo
@@ -107,23 +139,33 @@ install_centos() {
                    rocm-libs \
                    rccl \
                    rocprofiler-dev \
-                   roctracer-dev \
-                   amd-smi-lib
+                   roctracer-dev
 
   # precompiled miopen kernels; search for all unversioned packages
   # if search fails it will abort this script; use true to avoid case where search fails
-  MIOPENHIPGFX=$(yum -q search miopen-hip-gfx | grep miopen-hip-gfx | awk '{print $1}'| grep -F kdb. || true)
-  if [[ "x${MIOPENHIPGFX}" = x ]]; then
-    echo "miopen-hip-gfx package not available" && exit 1
+  if [[ $(ver $ROCM_VERSION) -ge $(ver 5.5) ]]; then
+      MIOPENHIPGFX=$(yum -q search miopen-hip-gfx | grep miopen-hip-gfx | awk '{print $1}'| grep -F kdb. || true)
+      if [[ "x${MIOPENHIPGFX}" = x ]]; then
+        echo "miopen-hip-gfx package not available" && exit 1
+      else
+        yum install -y ${MIOPENHIPGFX}
+      fi
   else
-    yum install -y ${MIOPENHIPGFX}
+      MIOPENKERNELS=$(yum -q search miopenkernels | grep miopenkernels- | awk '{print $1}'| grep -F kdb. || true)
+      if [[ "x${MIOPENKERNELS}" = x ]]; then
+        echo "miopenkernels package not available" && exit 1
+      else
+        yum install -y ${MIOPENKERNELS}
+      fi
   fi
 
   # ROCm 6.0 had a regression where journal_mode was enabled on the kdb files resulting in permission errors at runtime
-  for kdb in /opt/rocm/share/miopen/db/*.kdb
-  do
-      sqlite3 $kdb "PRAGMA journal_mode=off; PRAGMA VACUUM;"
-  done
+  if [[ $(ver $ROCM_VERSION) -ge $(ver 6.0) ]]; then
+      for kdb in /opt/rocm/share/miopen/db/*.kdb
+      do
+          sqlite3 $kdb "PRAGMA journal_mode=off; PRAGMA VACUUM;"
+      done
+  fi
 
   # Cleanup
   yum clean all

.ci/docker/common/install_triton.sh

@@ -13,11 +13,8 @@ conda_reinstall() {
 }
 
 if [ -n "${ROCM_VERSION}" ]; then
-  TRITON_REPO="https://github.com/openai/triton"
+  TRITON_REPO="https://github.com/ROCmSoftwarePlatform/triton"
   TRITON_TEXT_FILE="triton-rocm"
-elif [ -n "${XPU_VERSION}" ]; then
-  TRITON_REPO="https://github.com/intel/intel-xpu-backend-for-triton"
-  TRITON_TEXT_FILE="triton-xpu"
 else
   TRITON_REPO="https://github.com/openai/triton"
   TRITON_TEXT_FILE="triton"

.ci/docker/common/install_vision.sh

@@ -5,7 +5,8 @@ set -ex
 install_ubuntu() {
   apt-get update
   apt-get install -y --no-install-recommends \
-          libopencv-dev
+          libopencv-dev \
+          libavcodec-dev
 
   # Cleanup
   apt-get autoclean && apt-get clean
@@ -18,7 +19,8 @@ install_centos() {
   yum --enablerepo=extras install -y epel-release
 
   yum install -y \
-      opencv-devel
+      opencv-devel \
+      ffmpeg-devel
 
   # Cleanup
   yum clean all

.ci/docker/common/install_xpu.sh

@@ -3,7 +3,10 @@ set -xe
 
 
 # Intel® software for general purpose GPU capabilities.
-# Refer to https://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpus.html
+# Refer to https://dgpu-docs.intel.com/releases/stable_647_21_20230714.html
+
+# Intel® oneAPI Base Toolkit (version 2024.0.0) has been updated to include functional and security updates.
+# Refer to https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit-download.html
 
 # Users should update to the latest version as it becomes available
 
@@ -14,16 +17,14 @@ function install_ubuntu() {
     # Set up the repository. To do this, download the key to the system keyring
     wget -qO - https://repositories.intel.com/gpu/intel-graphics.key \
         | gpg --dearmor --output /usr/share/keyrings/intel-graphics.gpg
-    wget -qO - https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB \
-        | gpg --dearmor --output /usr/share/keyrings/intel-for-pytorch-gpu-dev-keyring.gpg
+    wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB \
+        | gpg --dearmor | tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null
 
     # Add the signed entry to APT sources and configure the APT client to use the Intel repository
-    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] \
-        https://repositories.intel.com/gpu/ubuntu jammy/lts/2350 unified" \
+    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy/production/2328 unified" \
         | tee /etc/apt/sources.list.d/intel-gpu-jammy.list
-    echo "deb [signed-by=/usr/share/keyrings/intel-for-pytorch-gpu-dev-keyring.gpg] \
-        https://apt.repos.intel.com/intel-for-pytorch-gpu-dev all main" \
-        | tee /etc/apt/sources.list.d/intel-for-pytorch-gpu-dev.list
+    echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" \
+        | tee /etc/apt/sources.list.d/oneAPI.list
 
     # Update the packages list and repository index
     apt-get update
@@ -39,11 +40,11 @@ function install_ubuntu() {
         mesa-vdpau-drivers mesa-vulkan-drivers va-driver-all vainfo hwinfo clinfo
     # Development Packages
     apt-get install -y libigc-dev intel-igc-cm libigdfcl-dev libigfxcmrt-dev level-zero-dev
-    # Install Intel Support Packages
-    if [ -n "$XPU_VERSION" ]; then
-        apt-get install -y intel-for-pytorch-gpu-dev-${XPU_VERSION}
+    # Install Intel® oneAPI Base Toolkit
+    if [ -n "$BASEKIT_VERSION" ]; then
+        apt-get install intel-basekit=$BASEKIT_VERSION -y
     else
-        apt-get install -y intel-for-pytorch-gpu-dev
+        apt-get install intel-basekit -y
     fi
 
     # Cleanup

.ci/docker/requirements-ci.txt

@@ -85,10 +85,10 @@ librosa>=0.6.2 ; python_version < "3.11"
 #Pinned versions:
 #test that import:
 
-mypy==1.10.0
+mypy==1.8.0
 # Pin MyPy version because new errors are likely to appear with each release
 #Description: linter
-#Pinned versions: 1.10.0
+#Pinned versions: 1.8.0
 #test that import: test_typing.py, test_type_hints.py
 
 networkx==2.8.8
@@ -134,9 +134,9 @@ opt-einsum==3.3
 #Pinned versions: 3.3
 #test that import: test_linalg.py
 
-optree==0.12.1
+optree==0.9.1
 #Description: A library for tree manipulation
-#Pinned versions: 0.12.1
+#Pinned versions: 0.9.1
 #test that import: test_vmap.py, test_aotdispatch.py, test_dynamic_shapes.py,
 #test_pytree.py, test_ops.py, test_control_flow.py, test_modules.py,
 #common_utils.py, test_eager_transforms.py, test_python_dispatch.py,
@@ -147,9 +147,9 @@ optree==0.12.1
 #test_pointwise_ops.py, test_dtensor_ops.py, test_torchinductor.py, test_fx.py,
 #test_fake_tensor.py, test_mps.py
 
-pillow==10.3.0
+pillow==10.2.0
 #Description:  Python Imaging Library fork
-#Pinned versions: 10.3.0
+#Pinned versions: 10.2.0
 #test that import:
 
 protobuf==3.20.2
@@ -228,11 +228,12 @@ scikit-image==0.20.0 ; python_version >= "3.10"
 #Pinned versions: 0.20.3
 #test that import:
 
-scipy==1.10.1 ; python_version <= "3.11"
-scipy==1.12.0 ; python_version == "3.12"
+scipy==1.6.3 ; python_version < "3.10"
+scipy==1.8.1 ; python_version == "3.10"
+scipy==1.10.1 ; python_version == "3.11"
 # Pin SciPy because of failing distribution tests (see #60347)
 #Description: scientific python
-#Pinned versions: 1.10.1
+#Pinned versions: 1.6.3
 #test that import: test_unary_ufuncs.py, test_torch.py,test_tensor_creation_ops.py
 #test_spectral_ops.py, test_sparse_csr.py, test_reductions.py,test_nn.py
 #test_linalg.py, test_binary_ufuncs.py
@@ -263,10 +264,10 @@ unittest-xml-reporting<=3.2.0,>=2.0.0
 #Pinned versions:
 #test that import:
 
-#lintrunner is supported on aarch64-linux only from 0.12.4 version
-lintrunner==0.12.5
+#wheel not found on aarch64, and source build requires rust
+lintrunner==0.10.7 ; platform_machine == "x86_64"
 #Description: all about linters!
-#Pinned versions: 0.12.5
+#Pinned versions: 0.10.7
 #test that import:
 
 rockset==1.0.3
@@ -279,9 +280,9 @@ ghstack==0.8.0
 #Pinned versions: 0.8.0
 #test that import:
 
-jinja2==3.1.4
+jinja2==3.1.3
 #Description: jinja2 template engine
-#Pinned versions: 3.1.4
+#Pinned versions: 3.1.3
 #test that import:
 
 pytest-cpp==2.3.0
@@ -306,9 +307,7 @@ pywavelets==1.5.0 ; python_version >= "3.12"
 #Pinned versions: 1.4.1
 #test that import:
 
-lxml==5.0.0
+lxml==5.0.0.
 #Description: This is a requirement of unittest-xml-reporting
 
 # Python-3.9 binaries
-
-PyGithub==2.3.0

.ci/docker/triton_version.txt

@@ -1 +1 @@
-3.0.0
+2.3.0

.ci/docker/ubuntu-cuda/Dockerfile

@@ -56,7 +56,7 @@ RUN if [ -n "${DB}" ]; then bash ./install_db.sh; fi
 RUN rm install_db.sh
 ENV INSTALLED_DB ${DB}
 
-# (optional) Install vision packages like OpenCV
+# (optional) Install vision packages like OpenCV and ffmpeg
 ARG VISION
 COPY ./common/install_vision.sh ./common/cache_vision_models.sh ./common/common_utils.sh ./
 RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
@@ -103,14 +103,6 @@ COPY triton_version.txt triton_version.txt
 RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi
 RUN rm install_triton.sh common_utils.sh triton.txt triton_version.txt
 
-ARG HALIDE
-# Build and install halide
-COPY ./common/install_halide.sh install_halide.sh
-COPY ./common/common_utils.sh common_utils.sh
-COPY ci_commit_pins/halide.txt halide.txt
-RUN if [ -n "${HALIDE}" ]; then bash ./install_halide.sh; fi
-RUN rm install_halide.sh common_utils.sh halide.txt
-
 # Install ccache/sccache (do this last, so we get priority in PATH)
 COPY ./common/install_cache.sh install_cache.sh
 ENV PATH /opt/cache/bin:$PATH
@@ -147,7 +139,7 @@ COPY --from=pytorch/llvm:9.0.1 /opt/llvm /opt/llvm
 ARG CUDNN_VERSION
 ARG CUDA_VERSION
 COPY ./common/install_cudnn.sh install_cudnn.sh
-RUN if [ -n "${CUDNN_VERSION}" ]; then bash install_cudnn.sh; fi
+RUN if [ "${CUDNN_VERSION}" -eq 8 ]; then bash install_cudnn.sh; fi
 RUN rm install_cudnn.sh
 
 # Install CUSPARSELT
@@ -160,7 +152,6 @@ RUN rm install_cusparselt.sh
 RUN if [ -h /usr/local/cuda-11.6/cuda-11.6 ]; then rm /usr/local/cuda-11.6/cuda-11.6; fi
 RUN if [ -h /usr/local/cuda-11.7/cuda-11.7 ]; then rm /usr/local/cuda-11.7/cuda-11.7; fi
 RUN if [ -h /usr/local/cuda-12.1/cuda-12.1 ]; then rm /usr/local/cuda-12.1/cuda-12.1; fi
-RUN if [ -h /usr/local/cuda-12.4/cuda-12.4 ]; then rm /usr/local/cuda-12.4/cuda-12.4; fi
 
 USER jenkins
 CMD ["bash"]

.ci/docker/ubuntu-rocm/Dockerfile

@@ -53,7 +53,7 @@ RUN if [ -n "${DB}" ]; then bash ./install_db.sh; fi
 RUN rm install_db.sh
 ENV INSTALLED_DB ${DB}
 
-# (optional) Install vision packages like OpenCV
+# (optional) Install vision packages like OpenCV and ffmpeg
 ARG VISION
 COPY ./common/install_vision.sh ./common/cache_vision_models.sh ./common/common_utils.sh ./
 RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
@@ -78,11 +78,6 @@ ENV MAGMA_HOME /opt/rocm/magma
 ENV LANG C.UTF-8
 ENV LC_ALL C.UTF-8
 
-# Install amdsmi
-COPY ./common/install_amdsmi.sh install_amdsmi.sh
-RUN bash ./install_amdsmi.sh
-RUN rm install_amdsmi.sh
-
 # (optional) Install non-default CMake version
 ARG CMAKE_VERSION
 COPY ./common/install_cmake.sh install_cmake.sh
@@ -105,13 +100,6 @@ COPY triton_version.txt triton_version.txt
 RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi
 RUN rm install_triton.sh common_utils.sh triton-rocm.txt triton_version.txt
 
-# Install AOTriton
-COPY ./aotriton_version.txt aotriton_version.txt
-COPY ./common/common_utils.sh common_utils.sh
-COPY ./common/install_aotriton.sh install_aotriton.sh
-RUN ["/bin/bash", "-c", "./install_aotriton.sh /opt/rocm && rm -rf install_aotriton.sh aotriton_version.txt common_utils.sh"]
-ENV AOTRITON_INSTALLED_PREFIX /opt/rocm/aotriton
-
 # Install ccache/sccache (do this last, so we get priority in PATH)
 COPY ./common/install_cache.sh install_cache.sh
 ENV PATH /opt/cache/bin:$PATH

.ci/docker/ubuntu-xpu/Dockerfile

@@ -61,20 +61,15 @@ COPY ci_commit_pins/timm.txt timm.txt
 RUN if [ -n "${INDUCTOR_BENCHMARKS}" ]; then bash ./install_inductor_benchmark_deps.sh; fi
 RUN rm install_inductor_benchmark_deps.sh common_utils.sh timm.txt huggingface.txt
 
-# Install XPU Dependencies
-ARG XPU_VERSION
-COPY ./common/install_xpu.sh install_xpu.sh
-RUN bash ./install_xpu.sh && rm install_xpu.sh
-
 ARG TRITON
 # Install triton, this needs to be done before sccache because the latter will
 # try to reach out to S3, which docker build runners don't have access
 COPY ./common/install_triton.sh install_triton.sh
 COPY ./common/common_utils.sh common_utils.sh
-COPY ci_commit_pins/triton-xpu.txt triton-xpu.txt
-COPY triton_version.txt triton_version.txt
+# TODO: will add triton xpu commit
+COPY ci_commit_pins/triton.txt triton.txt
 RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi
-RUN rm install_triton.sh common_utils.sh triton-xpu.txt triton_version.txt
+RUN rm install_triton.sh common_utils.sh triton.txt
 
 # (optional) Install database packages like LMDB and LevelDB
 ARG DB
@@ -83,13 +78,18 @@ RUN if [ -n "${DB}" ]; then bash ./install_db.sh; fi
 RUN rm install_db.sh
 ENV INSTALLED_DB ${DB}
 
-# (optional) Install vision packages like OpenCV
+# (optional) Install vision packages like OpenCV and ffmpeg
 ARG VISION
 COPY ./common/install_vision.sh ./common/cache_vision_models.sh ./common/common_utils.sh ./
 RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
 RUN rm install_vision.sh cache_vision_models.sh common_utils.sh
 ENV INSTALLED_VISION ${VISION}
 
+# Install XPU Dependencies
+ARG BASEKIT_VERSION
+COPY ./common/install_xpu.sh install_xpu.sh
+RUN bash ./install_xpu.sh && rm install_xpu.sh
+
 # (optional) Install non-default CMake version
 ARG CMAKE_VERSION
 COPY ./common/install_cmake.sh install_cmake.sh

.ci/docker/ubuntu/Dockerfile

@@ -80,7 +80,7 @@ RUN if [ -n "${DB}" ]; then bash ./install_db.sh; fi
 RUN rm install_db.sh
 ENV INSTALLED_DB ${DB}
 
-# (optional) Install vision packages like OpenCV
+# (optional) Install vision packages like OpenCV and ffmpeg
 ARG VISION
 COPY ./common/install_vision.sh ./common/cache_vision_models.sh ./common/common_utils.sh ./
 RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
@@ -155,14 +155,6 @@ COPY ci_commit_pins/executorch.txt executorch.txt
 RUN if [ -n "${EXECUTORCH}" ]; then bash ./install_executorch.sh; fi
 RUN rm install_executorch.sh common_utils.sh executorch.txt
 
-ARG HALIDE
-# Build and install halide
-COPY ./common/install_halide.sh install_halide.sh
-COPY ./common/common_utils.sh common_utils.sh
-COPY ci_commit_pins/halide.txt halide.txt
-RUN if [ -n "${HALIDE}" ]; then bash ./install_halide.sh; fi
-RUN rm install_halide.sh common_utils.sh halide.txt
-
 ARG ONNX
 # Install ONNX dependencies
 COPY ./common/install_onnx.sh ./common/common_utils.sh ./
@@ -177,11 +169,9 @@ RUN rm install_acl.sh
 ENV INSTALLED_ACL ${ACL}
 
 # Install ccache/sccache (do this last, so we get priority in PATH)
-ARG SKIP_SCCACHE_INSTALL
 COPY ./common/install_cache.sh install_cache.sh
 ENV PATH /opt/cache/bin:$PATH
-RUN if [ -z "${SKIP_SCCACHE_INSTALL}" ]; then bash ./install_cache.sh; fi
-RUN rm install_cache.sh
+RUN bash ./install_cache.sh && rm install_cache.sh
 
 # Add jni.h for java host build
 COPY ./common/install_jni.sh install_jni.sh
@@ -198,9 +188,7 @@ ARG BUILD_ENVIRONMENT
 ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
 
 # Install LLVM dev version (Defined in the pytorch/builder github repository)
-ARG SKIP_LLVM_SRC_BUILD_INSTALL
 COPY --from=pytorch/llvm:9.0.1 /opt/llvm /opt/llvm
-RUN if [ -n "${SKIP_LLVM_SRC_BUILD_INSTALL}" ]; then set -eu; rm -rf /opt/llvm; fi
 
 # AWS specific CUDA build guidance
 ENV TORCH_CUDA_ARCH_LIST Maxwell

.ci/onnx/common.sh

@@ -1,9 +1,5 @@
-#!/bin/bash
-
 set -ex
 
-source "$(dirname "${BASH_SOURCE[0]}")/../pytorch/common_utils.sh"
-
 LOCAL_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 ROOT_DIR=$(cd "$LOCAL_DIR"/../.. && pwd)
 TEST_DIR="$ROOT_DIR/test"

.ci/onnx/test.sh

@@ -3,20 +3,6 @@
 # shellcheck source=./common.sh
 source "$(dirname "${BASH_SOURCE[0]}")/common.sh"
 
-# Workaround for dind-rootless userid mapping (https://github.com/pytorch/ci-infra/issues/96)
-WORKSPACE_ORIGINAL_OWNER_ID=$(stat -c '%u' "/var/lib/jenkins/workspace")
-cleanup_workspace() {
-  echo "sudo may print the following warning message that can be ignored. The chown command will still run."
-  echo "    sudo: setrlimit(RLIMIT_STACK): Operation not permitted"
-  echo "For more details refer to https://github.com/sudo-project/sudo/issues/42"
-  sudo chown -R "$WORKSPACE_ORIGINAL_OWNER_ID" /var/lib/jenkins/workspace
-}
-# Disable shellcheck SC2064 as we want to parse the original owner immediately.
-# shellcheck disable=SC2064
-trap_add cleanup_workspace EXIT
-sudo chown -R jenkins /var/lib/jenkins/workspace
-git config --global --add safe.directory /var/lib/jenkins/workspace
-
 if [[ "$BUILD_ENVIRONMENT" == *onnx* ]]; then
   # TODO: This can be removed later once vision is also part of the Docker image
   pip install -q --user --no-use-pep517 "git+https://github.com/pytorch/vision.git@$(cat .github/ci_commit_pins/vision.txt)"

.ci/pytorch/build.sh

@@ -44,7 +44,15 @@ if [[ "$BUILD_ENVIRONMENT" == *cuda11* ]]; then
   fi
 fi
 
-if [[ ${BUILD_ENVIRONMENT} == *"parallelnative"* ]]; then
+if [[ ${BUILD_ENVIRONMENT} == *"caffe2"* ]]; then
+  echo "Caffe2 build is ON"
+  export BUILD_CAFFE2=ON
+fi
+
+if [[ ${BUILD_ENVIRONMENT} == *"paralleltbb"* ]]; then
+  export ATEN_THREADING=TBB
+  export USE_TBB=1
+elif [[ ${BUILD_ENVIRONMENT} == *"parallelnative"* ]]; then
   export ATEN_THREADING=NATIVE
 fi
 
@@ -73,22 +81,7 @@ if ! which conda; then
     export USE_MKLDNN=0
   fi
 else
-  # CMAKE_PREFIX_PATH precedences
-  # 1. $CONDA_PREFIX, if defined. This follows the pytorch official build instructions.
-  # 2. /opt/conda/envs/py_${ANACONDA_PYTHON_VERSION}, if ANACONDA_PYTHON_VERSION defined.
-  #    This is for CI, which defines ANACONDA_PYTHON_VERSION but not CONDA_PREFIX.
-  # 3. $(conda info --base). The fallback value of pytorch official build
-  #    instructions actually refers to this.
-  #    Commonly this is /opt/conda/
-  if [[ -v CONDA_PREFIX ]]; then
-    export CMAKE_PREFIX_PATH=${CONDA_PREFIX}
-  elif [[ -v ANACONDA_PYTHON_VERSION ]]; then
-    export CMAKE_PREFIX_PATH="/opt/conda/envs/py_${ANACONDA_PYTHON_VERSION}"
-  else
-    # already checked by `! which conda`
-    CMAKE_PREFIX_PATH="$(conda info --base)"
-    export CMAKE_PREFIX_PATH
-  fi
+  export CMAKE_PREFIX_PATH=/opt/conda
 
   # Workaround required for MKL library linkage
   # https://github.com/pytorch/pytorch/issues/119557
@@ -230,28 +223,6 @@ if [[ "${BUILD_ENVIRONMENT}" != *android* && "${BUILD_ENVIRONMENT}" != *cuda* ]]
   export BUILD_STATIC_RUNTIME_BENCHMARK=ON
 fi
 
-if [[ "$BUILD_ENVIRONMENT" == *-debug* ]]; then
-  export CMAKE_BUILD_TYPE=RelWithAssert
-fi
-
-# Do not change workspace permissions for ROCm CI jobs
-# as it can leave workspace with bad permissions for cancelled jobs
-if [[ "$BUILD_ENVIRONMENT" != *rocm* ]]; then
-  # Workaround for dind-rootless userid mapping (https://github.com/pytorch/ci-infra/issues/96)
-  WORKSPACE_ORIGINAL_OWNER_ID=$(stat -c '%u' "/var/lib/jenkins/workspace")
-  cleanup_workspace() {
-    echo "sudo may print the following warning message that can be ignored. The chown command will still run."
-    echo "    sudo: setrlimit(RLIMIT_STACK): Operation not permitted"
-    echo "For more details refer to https://github.com/sudo-project/sudo/issues/42"
-    sudo chown -R "$WORKSPACE_ORIGINAL_OWNER_ID" /var/lib/jenkins/workspace
-  }
-  # Disable shellcheck SC2064 as we want to parse the original owner immediately.
-  # shellcheck disable=SC2064
-  trap_add cleanup_workspace EXIT
-  sudo chown -R jenkins /var/lib/jenkins/workspace
-  git config --global --add safe.directory /var/lib/jenkins/workspace
-fi
-
 if [[ "$BUILD_ENVIRONMENT" == *-bazel-* ]]; then
   set -e
 
@@ -277,37 +248,16 @@ else
   ( ! get_exit_code python setup.py clean bad_argument )
 
   if [[ "$BUILD_ENVIRONMENT" != *libtorch* ]]; then
+
     # rocm builds fail when WERROR=1
     # XLA test build fails when WERROR=1
     # set only when building other architectures
     # or building non-XLA tests.
     if [[ "$BUILD_ENVIRONMENT" != *rocm*  &&
           "$BUILD_ENVIRONMENT" != *xla* ]]; then
-      if [[ "$BUILD_ENVIRONMENT" != *py3.8* ]]; then
-        # Install numpy-2.0 release candidate for builds
-        # Which should be backward compatible with Numpy-1.X
-        python -mpip install --pre numpy==2.0.0rc1
-      fi
-
-      WERROR=1 python setup.py clean
-
-      if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-        BUILD_LIBTORCH_WHL=1 BUILD_PYTHON_ONLY=0 python setup.py bdist_wheel
-        BUILD_LIBTORCH_WHL=0 BUILD_PYTHON_ONLY=1 python setup.py bdist_wheel --cmake
-      else
-        WERROR=1 python setup.py bdist_wheel
-      fi
+      WERROR=1 python setup.py bdist_wheel
     else
-      python setup.py clean
-      if [[ "$BUILD_ENVIRONMENT" == *xla* ]]; then
-        source .ci/pytorch/install_cache_xla.sh
-      fi
-      if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-        echo "USE_SPLIT_BUILD cannot be used with xla or rocm"
-        exit 1
-      else
-        python setup.py bdist_wheel
-      fi
+      python setup.py bdist_wheel
     fi
     pip_install_whl "$(echo dist/*.whl)"
 
@@ -346,10 +296,9 @@ else
     CUSTOM_OP_TEST="$PWD/test/custom_operator"
     python --version
     SITE_PACKAGES="$(python -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-
     mkdir -p "$CUSTOM_OP_BUILD"
     pushd "$CUSTOM_OP_BUILD"
-    cmake "$CUSTOM_OP_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch;$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
+    cmake "$CUSTOM_OP_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch" -DPYTHON_EXECUTABLE="$(which python)" \
           -DCMAKE_MODULE_PATH="$CUSTOM_TEST_MODULE_PATH" -DUSE_ROCM="$CUSTOM_TEST_USE_ROCM"
     make VERBOSE=1
     popd
@@ -362,7 +311,7 @@ else
     SITE_PACKAGES="$(python -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
     mkdir -p "$JIT_HOOK_BUILD"
     pushd "$JIT_HOOK_BUILD"
-    cmake "$JIT_HOOK_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch;$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
+    cmake "$JIT_HOOK_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch" -DPYTHON_EXECUTABLE="$(which python)" \
           -DCMAKE_MODULE_PATH="$CUSTOM_TEST_MODULE_PATH" -DUSE_ROCM="$CUSTOM_TEST_USE_ROCM"
     make VERBOSE=1
     popd
@@ -374,7 +323,7 @@ else
     python --version
     mkdir -p "$CUSTOM_BACKEND_BUILD"
     pushd "$CUSTOM_BACKEND_BUILD"
-    cmake "$CUSTOM_BACKEND_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch;$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
+    cmake "$CUSTOM_BACKEND_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch" -DPYTHON_EXECUTABLE="$(which python)" \
           -DCMAKE_MODULE_PATH="$CUSTOM_TEST_MODULE_PATH" -DUSE_ROCM="$CUSTOM_TEST_USE_ROCM"
     make VERBOSE=1
     popd
@@ -405,8 +354,4 @@ if [[ "$BUILD_ENVIRONMENT" != *libtorch* && "$BUILD_ENVIRONMENT" != *bazel* ]];
   python tools/stats/export_test_times.py
 fi
 
-# snadampal: skipping it till sccache support added for aarch64
-# https://github.com/pytorch/pytorch/issues/121559
-if [[ "$BUILD_ENVIRONMENT" != *aarch64* ]]; then
-  print_sccache_stats
-fi
+print_sccache_stats

.ci/pytorch/common_utils.sh

@@ -56,29 +56,9 @@ function assert_git_not_dirty() {
 function pip_install_whl() {
   # This is used to install PyTorch and other build artifacts wheel locally
   # without using any network connection
-
-  # Convert the input arguments into an array
-  local args=("$@")
-
-  # Check if the first argument contains multiple paths separated by spaces
-  if [[ "${args[0]}" == *" "* ]]; then
-    # Split the string by spaces into an array
-    IFS=' ' read -r -a paths <<< "${args[0]}"
-    # Loop through each path and install individually
-    for path in "${paths[@]}"; do
-      echo "Installing $path"
-      python3 -mpip install --no-index --no-deps "$path"
-    done
-  else
-    # Loop through each argument and install individually
-    for path in "${args[@]}"; do
-      echo "Installing $path"
-      python3 -mpip install --no-index --no-deps "$path"
-    done
-  fi
+  python3 -mpip install --no-index --no-deps "$@"
 }
 
-
 function pip_install() {
   # retry 3 times
   # old versions of pip don't have the "--progress-bar" flag
@@ -179,7 +159,7 @@ function install_torchvision() {
 }
 
 function install_tlparse() {
-  pip_install --user "tlparse==0.3.7"
+  pip_install --user "tlparse==0.3.5"
   PATH="$(python -m site --user-base)/bin:$PATH"
 }
 
@@ -198,7 +178,7 @@ function install_torchrec_and_fbgemm() {
 
 function clone_pytorch_xla() {
   if [[ ! -d ./xla ]]; then
-    git clone --recursive --quiet https://github.com/pytorch/xla.git
+    git clone --recursive -b r2.3 https://github.com/pytorch/xla.git
     pushd xla
     # pin the xla hash so that we don't get broken by changes to xla
     git checkout "$(cat ../.github/ci_commit_pins/xla.txt)"
@@ -208,6 +188,28 @@ function clone_pytorch_xla() {
   fi
 }
 
+function checkout_install_torchdeploy() {
+  local commit
+  commit=$(get_pinned_commit multipy)
+  pushd ..
+  git clone --recurse-submodules https://github.com/pytorch/multipy.git
+  pushd multipy
+  git checkout "${commit}"
+  python multipy/runtime/example/generate_examples.py
+  BUILD_CUDA_TESTS=1 pip install -e .
+  popd
+  popd
+}
+
+function test_torch_deploy(){
+ pushd ..
+ pushd multipy
+ ./multipy/runtime/build/test_deploy
+ ./multipy/runtime/build/test_deploy_gpu
+ popd
+ popd
+}
+
 function checkout_install_torchbench() {
   local commit
   commit=$(get_pinned_commit torchbench)
@@ -222,8 +224,6 @@ function checkout_install_torchbench() {
     # to install and test other models
     python install.py --continue_on_fail
   fi
-  echo "Print all dependencies after TorchBench is installed"
-  python -mpip freeze
   popd
 }
 

.ci/pytorch/create_test_cert.py

@@ -6,7 +6,6 @@ from cryptography.hazmat.primitives import hashes, serialization
 from cryptography.hazmat.primitives.asymmetric import rsa
 from cryptography.x509.oid import NameOID
 
-
 temp_dir = mkdtemp()
 print(temp_dir)
 

.ci/pytorch/docs-test.sh

@@ -6,4 +6,4 @@ source "$(dirname "${BASH_SOURCE[0]}")/common.sh"
 echo "Testing pytorch docs"
 
 cd docs
-TERM=vt100 make doctest
+make doctest

.ci/pytorch/install_cache_xla.sh

@@ -1,37 +0,0 @@
-#!/bin/bash
-
-# Script for installing sccache on the xla build job, which uses xla's docker
-# image and doesn't have sccache installed on it.  This is mostly copied from
-# .ci/docker/install_cache.sh.  Changes are: removing checks that will always
-# return the same thing, ex checks for for rocm, CUDA, and changing the path
-# where sccache is installed, and not changing /etc/environment.
-
-set -ex
-
-install_binary() {
-  echo "Downloading sccache binary from S3 repo"
-  curl --retry 3 https://s3.amazonaws.com/ossci-linux/sccache -o /tmp/cache/bin/sccache
-}
-
-mkdir -p /tmp/cache/bin
-mkdir -p /tmp/cache/lib
-export PATH="/tmp/cache/bin:$PATH"
-
-install_binary
-chmod a+x /tmp/cache/bin/sccache
-
-function write_sccache_stub() {
-  # Unset LD_PRELOAD for ps because of asan + ps issues
-  # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90589
-  # shellcheck disable=SC2086
-  # shellcheck disable=SC2059
-  printf "#!/bin/sh\nif [ \$(env -u LD_PRELOAD ps -p \$PPID -o comm=) != sccache ]; then\n  exec sccache $(which $1) \"\$@\"\nelse\n  exec $(which $1) \"\$@\"\nfi" > "/tmp/cache/bin/$1"
-  chmod a+x "/tmp/cache/bin/$1"
-}
-
-write_sccache_stub cc
-write_sccache_stub c++
-write_sccache_stub gcc
-write_sccache_stub g++
-write_sccache_stub clang
-write_sccache_stub clang++

.ci/pytorch/multigpu-test.sh

@@ -18,9 +18,7 @@ time python test/run_test.py --verbose -i distributed/test_c10d_gloo
 time python test/run_test.py --verbose -i distributed/test_c10d_nccl
 time python test/run_test.py --verbose -i distributed/test_c10d_spawn_gloo
 time python test/run_test.py --verbose -i distributed/test_c10d_spawn_nccl
-time python test/run_test.py --verbose -i distributed/test_compute_comm_reordering
 time python test/run_test.py --verbose -i distributed/test_store
-time python test/run_test.py --verbose -i distributed/test_symmetric_memory
 time python test/run_test.py --verbose -i distributed/test_pg_wrapper
 time python test/run_test.py --verbose -i distributed/rpc/cuda/test_tensorpipe_agent
 # FSDP tests
@@ -47,13 +45,6 @@ time python test/run_test.py --verbose -i distributed/test_device_mesh
 time python test/run_test.py --verbose -i distributed/tensor/parallel/test_ddp_2d_parallel
 time python test/run_test.py --verbose -i distributed/tensor/parallel/test_fsdp_2d_parallel
 time python test/run_test.py --verbose -i distributed/tensor/parallel/test_tp_examples
-time python test/run_test.py --verbose -i distributed/tensor/parallel/test_tp_random_state
-
-# FSDP2 tests
-time python test/run_test.py --verbose -i distributed/_composable/fsdp/test_fully_shard_training -- -k test_2d_mlp_with_nd_mesh
-
-# Pipelining composability tests
-time python test/run_test.py --verbose -i distributed/pipelining/test_composability.py
 
 # Other tests
 time python test/run_test.py --verbose -i test_cuda_primary_ctx

.ci/pytorch/perf_test/compare_with_baseline.py

@@ -3,7 +3,6 @@ import json
 import math
 import sys
 
-
 parser = argparse.ArgumentParser()
 parser.add_argument(
     "--test-name", dest="test_name", action="store", required=True, help="test name"
@@ -60,16 +59,16 @@ print("sample mean: ", sample_mean)
 print("sample sigma: ", sample_sigma)
 
 if math.isnan(sample_mean):
-    raise Exception("""Error: sample mean is NaN""")  # noqa: TRY002
+    raise Exception("""Error: sample mean is NaN""")
 elif math.isnan(sample_sigma):
-    raise Exception("""Error: sample sigma is NaN""")  # noqa: TRY002
+    raise Exception("""Error: sample sigma is NaN""")
 
 z_value = (sample_mean - mean) / sigma
 
 print("z-value: ", z_value)
 
 if z_value >= 3:
-    raise Exception(  # noqa: TRY002
+    raise Exception(
         f"""\n
 z-value >= 3, there is high chance of perf regression.\n
 To reproduce this regression, run

.ci/pytorch/perf_test/get_stats.py

@@ -3,7 +3,6 @@ import sys
 
 import numpy
 
-
 sample_data_list = sys.argv[1:]
 sample_data_list = [float(v.strip()) for v in sample_data_list]
 

.ci/pytorch/perf_test/update_commit_hash.py

@@ -1,7 +1,6 @@
 import json
 import sys
 
-
 data_file_path = sys.argv[1]
 commit_hash = sys.argv[2]
 

.ci/pytorch/print_sccache_log.py

@@ -1,6 +1,5 @@
 import sys
 
-
 log_file_path = sys.argv[1]
 
 with open(log_file_path) as f:

.ci/pytorch/python_doc_push_script.sh

@@ -26,8 +26,8 @@ echo "error: python_doc_push_script.sh: version (arg2) not specified"
 fi
 
 # Argument 1: Where to copy the built documentation to
-# (pytorch_docs/$install_path)
-install_path="${1:-${DOCS_INSTALL_PATH:-${DOCS_VERSION}}}"
+# (pytorch.github.io/$install_path)
+install_path="${1:-${DOCS_INSTALL_PATH:-docs/${DOCS_VERSION}}}"
 if [ -z "$install_path" ]; then
 echo "error: python_doc_push_script.sh: install_path (arg1) not specified"
   exit 1
@@ -68,8 +68,8 @@ build_docs () {
 }
 
 
-git clone https://github.com/pytorch/docs pytorch_docs -b "$branch" --depth 1
-pushd pytorch_docs
+git clone https://github.com/pytorch/pytorch.github.io -b "$branch" --depth 1
+pushd pytorch.github.io
 
 export LC_ALL=C
 export PATH=/opt/conda/bin:$PATH
@@ -105,7 +105,6 @@ if [ "$is_main_doc" = true ]; then
     echo undocumented objects found:
     cat build/coverage/python.txt
     echo "Make sure you've updated relevant .rsts in docs/source!"
-    echo "You can reproduce locally by running 'cd docs && make coverage && cat build/coverage/python.txt'"
     exit 1
   fi
 else

.ci/pytorch/test.sh

@@ -6,27 +6,6 @@
 
 set -ex
 
-# shellcheck source=./common.sh
-source "$(dirname "${BASH_SOURCE[0]}")/common.sh"
-
-# Do not change workspace permissions for ROCm CI jobs
-# as it can leave workspace with bad permissions for cancelled jobs
-if [[ "$BUILD_ENVIRONMENT" != *rocm* ]]; then
-  # Workaround for dind-rootless userid mapping (https://github.com/pytorch/ci-infra/issues/96)
-  WORKSPACE_ORIGINAL_OWNER_ID=$(stat -c '%u' "/var/lib/jenkins/workspace")
-  cleanup_workspace() {
-    echo "sudo may print the following warning message that can be ignored. The chown command will still run."
-    echo "    sudo: setrlimit(RLIMIT_STACK): Operation not permitted"
-    echo "For more details refer to https://github.com/sudo-project/sudo/issues/42"
-    sudo chown -R "$WORKSPACE_ORIGINAL_OWNER_ID" /var/lib/jenkins/workspace
-  }
-  # Disable shellcheck SC2064 as we want to parse the original owner immediately.
-  # shellcheck disable=SC2064
-  trap_add cleanup_workspace EXIT
-  sudo chown -R jenkins /var/lib/jenkins/workspace
-  git config --global --add safe.directory /var/lib/jenkins/workspace
-fi
-
 echo "Environment variables:"
 env
 
@@ -111,6 +90,9 @@ if [[ -n $TESTS_TO_INCLUDE ]]; then
   INCLUDE_CLAUSE="--include $TESTS_TO_INCLUDE"
 fi
 
+# shellcheck source=./common.sh
+source "$(dirname "${BASH_SOURCE[0]}")/common.sh"
+
 echo "Environment variables"
 env
 
@@ -181,11 +163,6 @@ if [[ "$BUILD_ENVIRONMENT" != *-bazel-* ]] ; then
   export PATH="$HOME/.local/bin:$PATH"
 fi
 
-if [[ "$BUILD_ENVIRONMENT" == *aarch64* ]]; then
-  # TODO: revisit this once the CI is stabilized on aarch64 linux
-  export VALGRIND=OFF
-fi
-
 install_tlparse
 
 # DANGER WILL ROBINSON.  The LD_PRELOAD here could cause you problems
@@ -234,6 +211,8 @@ if [[ "$BUILD_ENVIRONMENT" == *asan* ]]; then
     export LD_PRELOAD=/usr/lib/llvm-15/lib/clang/15.0.7/lib/linux/libclang_rt.asan-x86_64.so
     # Disable valgrind for asan
     export VALGRIND=OFF
+    # Increase stack size, because ASAN red zones use more stack
+    ulimit -s 81920
 
     (cd test && python -c "import torch; print(torch.__version__, torch.version.git_version)")
     echo "The next four invocations are expected to crash; if they don't that means ASAN/UBSAN is misconfigured"
@@ -249,7 +228,9 @@ fi
 # This tests that the debug asserts are working correctly.
 if [[ "$BUILD_ENVIRONMENT" == *-debug* ]]; then
     echo "We are in debug mode: $BUILD_ENVIRONMENT. Expect the python assertion to fail"
-    (cd test && ! get_exit_code python -c "import torch; torch._C._crash_if_debug_asserts_fail(424242)")
+    # TODO: Enable the check after we setup the build to run debug asserts without having
+    #       to do a full (and slow) debug build
+    # (cd test && ! get_exit_code python -c "import torch; torch._C._crash_if_debug_asserts_fail(424242)")
 elif [[ "$BUILD_ENVIRONMENT" != *-bazel-* ]]; then
     # Noop when debug is disabled. Skip bazel jobs because torch isn't available there yet.
     echo "We are not in debug mode: $BUILD_ENVIRONMENT. Expect the assertion to pass"
@@ -275,9 +256,6 @@ test_python_shard() {
 
   # Bare --include flag is not supported and quoting for lint ends up with flag not being interpreted correctly
   # shellcheck disable=SC2086
-
-  # modify LD_LIBRARY_PATH to ensure it has the conda env.
-  # This set of tests has been shown to be buggy without it for the split-build
   time python test/run_test.py --exclude-jit-executor --exclude-distributed-tests $INCLUDE_CLAUSE --shard "$1" "$NUM_TEST_SHARDS" --verbose $PYTHON_TEST_EXTRA_OPTION
 
   assert_git_not_dirty
@@ -311,25 +289,19 @@ test_dynamo_shard() {
 test_inductor_distributed() {
   # Smuggle a few multi-gpu tests here so that we don't have to request another large node
   echo "Testing multi_gpu tests in test_torchinductor"
-  python test/run_test.py -i inductor/test_torchinductor.py -k test_multi_gpu --verbose
-  python test/run_test.py -i inductor/test_aot_inductor.py -k test_non_default_cuda_device --verbose
-  python test/run_test.py -i inductor/test_aot_inductor.py -k test_replicate_on_devices --verbose
-  python test/run_test.py -i distributed/test_c10d_functional_native.py --verbose
-  python test/run_test.py -i distributed/_tensor/test_dtensor_compile.py --verbose
-  python test/run_test.py -i distributed/tensor/parallel/test_fsdp_2d_parallel.py --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_comm.py --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_multi_group --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_with_activation_checkpointing --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_2d_mlp --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_hsdp --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_2d_transformer_checkpoint_resume --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_gradient_accumulation --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_state_dict.py -k test_dp_state_dict_save_load --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_frozen.py --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_mixed_precision.py -k test_compute_dtype --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_mixed_precision.py -k test_reduce_dtype --verbose
-  python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_clip_grad_norm_.py -k test_clip_grad_norm_2d --verbose
-  python test/run_test.py -i distributed/fsdp/test_fsdp_tp_integration.py -k test_fsdp_tp_integration --verbose
+  pytest test/inductor/test_torchinductor.py -k test_multi_gpu
+  pytest test/inductor/test_aot_inductor.py -k test_non_default_cuda_device
+  pytest test/inductor/test_aot_inductor.py -k test_replicate_on_devices
+  pytest test/distributed/test_c10d_functional_native.py
+  pytest test/distributed/_tensor/test_dtensor_compile.py
+  pytest test/distributed/tensor/parallel/test_fsdp_2d_parallel.py
+  pytest test/distributed/_composable/fsdp/test_fully_shard_comm.py
+  pytest test/distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_multi_group
+  pytest test/distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_with_activation_checkpointing
+  pytest test/distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_2d_mlp
+  pytest test/distributed/_composable/fsdp/test_fully_shard_frozen.py
+  pytest test/distributed/_composable/fsdp/test_fully_shard_mixed_precision.py -k test_compute_dtype
+  pytest test/distributed/_composable/fsdp/test_fully_shard_mixed_precision.py -k test_reduce_dtype
 
   # this runs on both single-gpu and multi-gpu instance. It should be smart about skipping tests that aren't supported
   # with if required # gpus aren't available
@@ -337,50 +309,26 @@ test_inductor_distributed() {
   assert_git_not_dirty
 }
 
-test_inductor_shard() {
-  if [[ -z "$NUM_TEST_SHARDS" ]]; then
-    echo "NUM_TEST_SHARDS must be defined to run a Python test shard"
-    exit 1
-  fi
-
+test_inductor() {
   python tools/dynamo/verify_dynamo.py
-  python test/run_test.py --inductor \
-    --include test_modules test_ops test_ops_gradients test_torch \
-    --shard "$1" "$NUM_TEST_SHARDS" \
-    --verbose
-
+  python test/run_test.py --inductor --include test_modules test_ops test_ops_gradients test_torch --verbose
   # Do not add --inductor for the following inductor unit tests, otherwise we will fail because of nested dynamo state
-  python test/run_test.py \
-    --include inductor/test_torchinductor inductor/test_torchinductor_opinfo inductor/test_aot_inductor \
-    --shard "$1" "$NUM_TEST_SHARDS" \
-    --verbose
-}
+  python test/run_test.py --include inductor/test_torchinductor inductor/test_torchinductor_opinfo --verbose
 
-test_inductor_aoti() {
   # docker build uses bdist_wheel which does not work with test_aot_inductor
   # TODO: need a faster way to build
   if [[ "$BUILD_ENVIRONMENT" != *rocm* ]]; then
-    BUILD_AOT_INDUCTOR_TEST=1 python setup.py develop
-    CPP_TESTS_DIR="${BUILD_BIN_DIR}" LD_LIBRARY_PATH="${TORCH_LIB_DIR}" python test/run_test.py --cpp --verbose -i cpp/test_aoti_abi_check cpp/test_aoti_inference
+      BUILD_AOT_INDUCTOR_TEST=1 python setup.py develop
+      CPP_TESTS_DIR="${BUILD_BIN_DIR}" LD_LIBRARY_PATH="${TORCH_LIB_DIR}" python test/run_test.py --cpp --verbose -i cpp/test_aot_inductor
   fi
 }
 
 test_inductor_cpp_wrapper_abi_compatible() {
   export TORCHINDUCTOR_ABI_COMPATIBLE=1
-  TEST_REPORTS_DIR=$(pwd)/test/test-reports
-  mkdir -p "$TEST_REPORTS_DIR"
-
   echo "Testing Inductor cpp wrapper mode with TORCHINDUCTOR_ABI_COMPATIBLE=1"
   # cpu stack allocation causes segfault and needs more investigation
-  PYTORCH_TESTING_DEVICE_ONLY_FOR="" python test/run_test.py --include inductor/test_cpu_cpp_wrapper
+  TORCHINDUCTOR_STACK_ALLOCATION=0 python test/run_test.py --include inductor/test_cpu_cpp_wrapper
   python test/run_test.py --include inductor/test_cuda_cpp_wrapper
-
-  TORCHINDUCTOR_CPP_WRAPPER=1 python benchmarks/dynamo/timm_models.py --device cuda --accuracy --amp \
-    --training --inductor --disable-cudagraphs --only vit_base_patch16_224 \
-    --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_training.csv"
-  python benchmarks/dynamo/check_accuracy.py \
-    --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_training.csv" \
-    --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_timm_training.csv"
 }
 
 # "Global" flags for inductor benchmarking controlled by TEST_CONFIG
@@ -405,7 +353,7 @@ if [[ "${TEST_CONFIG}" == *dynamic* ]]; then
   DYNAMO_BENCHMARK_FLAGS+=(--dynamic-shapes --dynamic-batch-only)
 fi
 
-if [[ "${TEST_CONFIG}" == *cpu_inductor* || "${TEST_CONFIG}" == *cpu_aot_inductor* ]]; then
+if [[ "${TEST_CONFIG}" == *cpu_inductor* ]]; then
   DYNAMO_BENCHMARK_FLAGS+=(--device cpu)
 else
   DYNAMO_BENCHMARK_FLAGS+=(--device cuda)
@@ -484,17 +432,6 @@ test_perf_for_dashboard() {
             "${target_flag[@]}" --"$mode" --"$dtype" --backend "$backend" "$@" \
             --output "$TEST_REPORTS_DIR/${backend}_max_autotune_${suite}_${dtype}_${mode}_cuda_${target}.csv"
       fi
-      if [[ "$DASHBOARD_TAG" == *cudagraphs_low_precision-true* ]] && [[ "$mode" == "inference" ]]; then
-        # TODO: This has a new dtype called quant and the benchmarks script needs to be updated to support this.
-        # The tentative command is as follows. It doesn't work now, but it's ok because we only need mock data
-        # to fill the dashboard.
-        python "benchmarks/dynamo/$suite.py" \
-          "${target_flag[@]}" --"$mode" --quant --backend "$backend" "$@" \
-          --output "$TEST_REPORTS_DIR/${backend}_cudagraphs_low_precision_${suite}_quant_${mode}_cuda_${target}.csv" || true
-        # Copy cudagraph results as mock data, easiest choice?
-        cp "$TEST_REPORTS_DIR/${backend}_with_cudagraphs_${suite}_${dtype}_${mode}_cuda_${target}.csv" \
-          "$TEST_REPORTS_DIR/${backend}_cudagraphs_low_precision_${suite}_quant_${mode}_cuda_${target}.csv"
-      fi
     done
   done
 }
@@ -530,10 +467,9 @@ test_single_dynamo_benchmark() {
     test_perf_for_dashboard "$suite" \
       "${DYNAMO_BENCHMARK_FLAGS[@]}" "$@" "${partition_flags[@]}"
   else
-    if [[ "${TEST_CONFIG}" == *aot_inductor* && "${TEST_CONFIG}" != *cpu_aot_inductor* ]]; then
+    if [[ "${TEST_CONFIG}" == *aot_inductor* ]]; then
       # Test AOTInductor with the ABI-compatible mode on CI
       # This can be removed once the ABI-compatible mode becomes default.
-      # For CPU device, we perfer non ABI-compatible mode on CI when testing AOTInductor.
       export TORCHINDUCTOR_ABI_COMPATIBLE=1
     fi
     python "benchmarks/dynamo/$suite.py" \
@@ -550,16 +486,6 @@ test_single_dynamo_benchmark() {
   fi
 }
 
-test_inductor_micro_benchmark() {
-  TEST_REPORTS_DIR=$(pwd)/test/test-reports
-  python benchmarks/gpt_fast/benchmark.py --output "${TEST_REPORTS_DIR}/gpt_fast_benchmark.csv"
-}
-
-test_inductor_halide() {
-  python test/run_test.py --include inductor/test_halide.py --verbose
-  assert_git_not_dirty
-}
-
 test_dynamo_benchmark() {
   # Usage: test_dynamo_benchmark huggingface 0
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
@@ -574,16 +500,8 @@ test_dynamo_benchmark() {
   elif [[ "${TEST_CONFIG}" == *perf* ]]; then
     test_single_dynamo_benchmark "dashboard" "$suite" "$shard_id" "$@"
   else
-    if [[ "${TEST_CONFIG}" == *cpu_inductor* || "${TEST_CONFIG}" == *cpu_aot_inductor* ]]; then
-      local dt="float32"
-      if [[ "${TEST_CONFIG}" == *amp* ]]; then
-        dt="amp"
-      fi
-      if [[ "${TEST_CONFIG}" == *freezing* ]]; then
-        test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --"$dt" --freezing "$@"
-      else
-        test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --"$dt" "$@"
-      fi
+    if [[ "${TEST_CONFIG}" == *cpu_inductor* ]]; then
+      test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --float32 "$@"
     elif [[ "${TEST_CONFIG}" == *aot_inductor* ]]; then
       test_single_dynamo_benchmark "inference" "$suite" "$shard_id" --inference --bfloat16 "$@"
     else
@@ -597,16 +515,12 @@ test_inductor_torchbench_smoketest_perf() {
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
   mkdir -p "$TEST_REPORTS_DIR"
 
-  # Test some models in the cpp wrapper mode
-  TORCHINDUCTOR_ABI_COMPATIBLE=1 TORCHINDUCTOR_CPP_WRAPPER=1 python benchmarks/dynamo/torchbench.py --device cuda --accuracy \
-    --bfloat16 --inference --inductor --only hf_T5 --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
-  TORCHINDUCTOR_ABI_COMPATIBLE=1 TORCHINDUCTOR_CPP_WRAPPER=1 python benchmarks/dynamo/torchbench.py --device cuda --accuracy \
-    --bfloat16 --inference --inductor --only llama --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
-  TORCHINDUCTOR_ABI_COMPATIBLE=1 TORCHINDUCTOR_CPP_WRAPPER=1 python benchmarks/dynamo/torchbench.py --device cuda --accuracy \
-    --bfloat16 --inference --inductor --only moco --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
+  # smoke test the cpp_wrapper mode
+  TORCHINDUCTOR_CPP_WRAPPER=1 python benchmarks/dynamo/torchbench.py --device cuda --accuracy --bfloat16 \
+    --inference --inductor --only hf_T5 --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_smoketest.csv"
   python benchmarks/dynamo/check_accuracy.py \
-    --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" \
-    --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_inference.csv"
+      --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_smoketest.csv" \
+      --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_inference.csv"
 
   python benchmarks/dynamo/torchbench.py --device cuda --performance --backend inductor --float16 --training \
     --batch-size-file "$(realpath benchmarks/dynamo/torchbench_models_list.txt)" --only hf_Bert \
@@ -621,8 +535,7 @@ test_inductor_torchbench_smoketest_perf() {
   # https://github.com/pytorch/pytorch/actions/runs/7158691360/job/19491437314,
   # and thus we lower its threshold to reduce flakiness. If this continues to be a problem,
   # we switch to use some other model.
-  # lowering threshold from 4.9 to 4.7 for cu124. Will bump it up after cuda 12.4.0->12.4.1 update
-  python benchmarks/dynamo/check_perf_csv.py -f "$TEST_REPORTS_DIR/inductor_inference_smoketest.csv" -t 4.7
+  python benchmarks/dynamo/check_perf_csv.py -f "$TEST_REPORTS_DIR/inductor_inference_smoketest.csv" -t 4.9
 
   # Check memory compression ratio for a few models
   for test in hf_Albert timm_vision_transformer; do
@@ -634,15 +547,6 @@ test_inductor_torchbench_smoketest_perf() {
       "$TEST_REPORTS_DIR/inductor_training_smoketest_$test.csv" \
       --expected benchmarks/dynamo/expected_ci_perf_inductor_torchbench.csv
   done
-
-  # Perform some "warm-start" runs for a few huggingface models.
-  for test in AlbertForQuestionAnswering AllenaiLongformerBase DistilBertForMaskedLM DistillGPT2 GoogleFnet YituTechConvBert; do
-    python benchmarks/dynamo/huggingface.py --accuracy --training --amp --inductor --device cuda --warm-start-latency \
-      --only $test --output "$TEST_REPORTS_DIR/inductor_warm_start_smoketest_$test.csv"
-    python benchmarks/dynamo/check_accuracy.py \
-      --actual "$TEST_REPORTS_DIR/inductor_warm_start_smoketest_$test.csv" \
-      --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_huggingface_training.csv"
-  done
 }
 
 test_inductor_torchbench_cpu_smoketest_perf(){
@@ -689,12 +593,6 @@ test_inductor_torchbench_cpu_smoketest_perf(){
   done
 }
 
-test_torchbench_gcp_smoketest(){
-  pushd "${TORCHBENCHPATH}"
-  python test.py -v
-  popd
-}
-
 test_python_gloo_with_tls() {
   source "$(dirname "${BASH_SOURCE[0]}")/run_glootls_test.sh"
   assert_git_not_dirty
@@ -726,6 +624,7 @@ test_aten() {
   ${SUDO} ln -sf "$TORCH_LIB_DIR"/libmkldnn* "$TEST_BASE_DIR"
   ${SUDO} ln -sf "$TORCH_LIB_DIR"/libnccl* "$TEST_BASE_DIR"
   ${SUDO} ln -sf "$TORCH_LIB_DIR"/libtorch* "$TEST_BASE_DIR"
+  ${SUDO} ln -sf "$TORCH_LIB_DIR"/libtbb* "$TEST_BASE_DIR"
 
   ls "$TEST_BASE_DIR"
   aten/tools/run_tests.sh "$TEST_BASE_DIR"
@@ -750,6 +649,21 @@ test_without_numpy() {
   popd
 }
 
+# pytorch extensions require including torch/extension.h which includes all.h
+# which includes utils.h which includes Parallel.h.
+# So you can call for instance parallel_for() from your extension,
+# but the compilation will fail because of Parallel.h has only declarations
+# and definitions are conditionally included Parallel.h(see last lines of Parallel.h).
+# I tried to solve it #39612 and #39881 by including Config.h into Parallel.h
+# But if Pytorch is built with TBB it provides Config.h
+# that has AT_PARALLEL_NATIVE_TBB=1(see #3961 or #39881) and it means that if you include
+# torch/extension.h which transitively includes Parallel.h
+# which transitively includes tbb.h which is not available!
+if [[ "${BUILD_ENVIRONMENT}" == *tbb* ]]; then
+  sudo mkdir -p /usr/include/tbb
+  sudo cp -r "$PWD"/third_party/tbb/include/tbb/* /usr/include/tbb
+fi
+
 test_libtorch() {
   local SHARD="$1"
 
@@ -763,6 +677,7 @@ test_libtorch() {
     ln -sf "$TORCH_LIB_DIR"/libc10* "$TORCH_BIN_DIR"
     ln -sf "$TORCH_LIB_DIR"/libshm* "$TORCH_BIN_DIR"
     ln -sf "$TORCH_LIB_DIR"/libtorch* "$TORCH_BIN_DIR"
+    ln -sf "$TORCH_LIB_DIR"/libtbb* "$TORCH_BIN_DIR"
     ln -sf "$TORCH_LIB_DIR"/libnvfuser* "$TORCH_BIN_DIR"
 
     export CPP_TESTS_DIR="${TORCH_BIN_DIR}"
@@ -899,6 +814,7 @@ test_rpc() {
   # test reporting process to function as expected.
   ln -sf "$TORCH_LIB_DIR"/libtorch* "$TORCH_BIN_DIR"
   ln -sf "$TORCH_LIB_DIR"/libc10* "$TORCH_BIN_DIR"
+  ln -sf "$TORCH_LIB_DIR"/libtbb* "$TORCH_BIN_DIR"
 
   CPP_TESTS_DIR="${TORCH_BIN_DIR}" python test/run_test.py --cpp --verbose -i cpp/test_cpp_rpc
 }
@@ -1178,21 +1094,15 @@ test_executorch() {
 
   pushd /executorch
 
-  export PYTHON_EXECUTABLE=python
-  export EXECUTORCH_BUILD_PYBIND=ON
-  export CMAKE_ARGS="-DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON"
-
-  # NB: We need to rebuild ExecuTorch runner here because it depends on PyTorch
-  # from the PR
+  # NB: We need to build ExecuTorch runner here and not inside the Docker image
+  # because it depends on PyTorch
   # shellcheck disable=SC1091
-  source .ci/scripts/setup-linux.sh cmake
-
-  echo "Run ExecuTorch unit tests"
-  pytest -v -n auto
-  # shellcheck disable=SC1091
-  LLVM_PROFDATA=llvm-profdata-12 LLVM_COV=llvm-cov-12 bash test/run_oss_cpp_tests.sh
+  source .ci/scripts/utils.sh
+  build_executorch_runner "cmake"
 
   echo "Run ExecuTorch regression tests for some models"
+  # NB: This is a sample model, more can be added here
+  export PYTHON_EXECUTABLE=python
   # TODO(huydhn): Add more coverage here using ExecuTorch's gather models script
   # shellcheck disable=SC1091
   source .ci/scripts/test.sh mv3 cmake xnnpack-quantization-delegation ''
@@ -1206,33 +1116,11 @@ test_executorch() {
   assert_git_not_dirty
 }
 
-test_linux_aarch64(){
-  python test/run_test.py --include test_modules test_mkldnn test_mkldnn_fusion test_openmp test_torch test_dynamic_shapes \
-       test_transformers test_multiprocessing test_numpy_interop --verbose
-
-  # Dynamo tests
-  python test/run_test.py --include dynamo/test_compile dynamo/test_backends dynamo/test_comptime dynamo/test_config \
-       dynamo/test_functions dynamo/test_fx_passes_pre_grad dynamo/test_interop dynamo/test_model_output dynamo/test_modules \
-       dynamo/test_optimizers dynamo/test_recompile_ux dynamo/test_recompiles --verbose
-
-  # Inductor tests
-  python test/run_test.py --include inductor/test_torchinductor inductor/test_benchmark_fusion inductor/test_codecache \
-       inductor/test_config inductor/test_control_flow inductor/test_coordinate_descent_tuner inductor/test_fx_fusion \
-       inductor/test_group_batch_fusion inductor/test_inductor_freezing inductor/test_inductor_utils \
-       inductor/test_inplacing_pass inductor/test_kernel_benchmark inductor/test_layout_optim \
-       inductor/test_max_autotune inductor/test_memory_planning inductor/test_metrics inductor/test_multi_kernel inductor/test_pad_mm \
-       inductor/test_pattern_matcher inductor/test_perf inductor/test_profiler inductor/test_select_algorithm inductor/test_smoke \
-       inductor/test_split_cat_fx_passes inductor/test_standalone_compile inductor/test_torchinductor \
-       inductor/test_torchinductor_codegen_dynamic_shapes inductor/test_torchinductor_dynamic_shapes --verbose
-}
-
 if ! [[ "${BUILD_ENVIRONMENT}" == *libtorch* || "${BUILD_ENVIRONMENT}" == *-bazel-* ]]; then
   (cd test && python -c "import torch; print(torch.__config__.show())")
   (cd test && python -c "import torch; print(torch.__config__.parallel_info())")
 fi
-if [[ "$BUILD_ENVIRONMENT" == *aarch64* ]]; then
-  test_linux_aarch64
-elif [[ "${TEST_CONFIG}" == *backward* ]]; then
+if [[ "${TEST_CONFIG}" == *backward* ]]; then
   test_forward_backward_compatibility
   # Do NOT add tests after bc check tests, see its comment.
 elif [[ "${TEST_CONFIG}" == *xla* ]]; then
@@ -1252,12 +1140,11 @@ elif [[ "$TEST_CONFIG" == distributed ]]; then
   if [[ "${SHARD_NUMBER}" == 1 ]]; then
     test_rpc
   fi
+elif [[ "$TEST_CONFIG" == deploy ]]; then
+  checkout_install_torchdeploy
+  test_torch_deploy
 elif [[ "${TEST_CONFIG}" == *inductor_distributed* ]]; then
   test_inductor_distributed
-elif [[ "${TEST_CONFIG}" == *inductor-halide* ]]; then
-  test_inductor_halide
-elif [[ "${TEST_CONFIG}" == *inductor-micro-benchmark* ]]; then
-  test_inductor_micro_benchmark
 elif [[ "${TEST_CONFIG}" == *huggingface* ]]; then
   install_torchvision
   id=$((SHARD_NUMBER-1))
@@ -1267,14 +1154,13 @@ elif [[ "${TEST_CONFIG}" == *timm* ]]; then
   id=$((SHARD_NUMBER-1))
   test_dynamo_benchmark timm_models "$id"
 elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
-  if [[ "${TEST_CONFIG}" == *cpu_inductor* || "${TEST_CONFIG}" == *cpu_aot_inductor* ]]; then
+  if [[ "${TEST_CONFIG}" == *cpu_inductor* ]]; then
     install_torchaudio cpu
   else
     install_torchaudio cuda
   fi
   install_torchtext
   install_torchvision
-  TORCH_CUDA_ARCH_LIST="8.0;8.6" pip_install git+https://github.com/pytorch/ao.git
   id=$((SHARD_NUMBER-1))
   # https://github.com/opencv/opencv-python/issues/885
   pip_install opencv-python==4.8.0.74
@@ -1286,14 +1172,11 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
       llama_v2_7b_16h resnet50 timm_efficientnet mobilenet_v3_large timm_resnest \
       shufflenet_v2_x1_0 hf_GPT2
     PYTHONPATH=$(pwd)/torchbench test_inductor_torchbench_cpu_smoketest_perf
-  elif [[ "${TEST_CONFIG}" == *torchbench_gcp_smoketest* ]]; then
-    checkout_install_torchbench
-    TORCHBENCHPATH=$(pwd)/torchbench test_torchbench_gcp_smoketest
   else
     checkout_install_torchbench
     # Do this after checkout_install_torchbench to ensure we clobber any
     # nightlies that torchbench may pull in
-    if [[ "${TEST_CONFIG}" != *cpu_inductor* && "${TEST_CONFIG}" != *cpu_aot_inductor* ]]; then
+    if [[ "${TEST_CONFIG}" != *cpu_inductor* ]]; then
       install_torchrec_and_fbgemm
     fi
     PYTHONPATH=$(pwd)/torchbench test_dynamo_benchmark torchbench "$id"
@@ -1301,23 +1184,17 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
 elif [[ "${TEST_CONFIG}" == *inductor_cpp_wrapper_abi_compatible* ]]; then
   install_torchvision
   test_inductor_cpp_wrapper_abi_compatible
-elif [[ "${TEST_CONFIG}" == *inductor* ]]; then
+elif [[ "${TEST_CONFIG}" == *inductor* && "${SHARD_NUMBER}" == 1 ]]; then
   install_torchvision
-  test_inductor_shard "${SHARD_NUMBER}"
-  if [[ "${SHARD_NUMBER}" == 1 ]]; then
-    test_inductor_aoti
-    test_inductor_distributed
-  fi
-elif [[ "${TEST_CONFIG}" == *dynamo* ]]; then
+  test_inductor
+  test_inductor_distributed
+elif [[ "${TEST_CONFIG}" == *dynamo* && "${SHARD_NUMBER}" == 1 && $NUM_TEST_SHARDS -gt 1 ]]; then
+  install_torchvision
+  test_dynamo_shard 1
+  test_aten
+elif [[ "${TEST_CONFIG}" == *dynamo* && $SHARD_NUMBER -gt 1 && $NUM_TEST_SHARDS -gt 1 ]]; then
   install_torchvision
   test_dynamo_shard "${SHARD_NUMBER}"
-  if [[ "${SHARD_NUMBER}" == 1 ]]; then
-    test_aten
-  fi
-elif [[ "${BUILD_ENVIRONMENT}" == *rocm* && -n "$TESTS_TO_INCLUDE" ]]; then
-  install_torchvision
-  test_python_shard "$SHARD_NUMBER"
-  test_aten
 elif [[ "${SHARD_NUMBER}" == 1 && $NUM_TEST_SHARDS -gt 1 ]]; then
   test_without_numpy
   install_torchvision
@@ -1347,6 +1224,10 @@ elif [[ "${BUILD_ENVIRONMENT}" == *-mobile-lightweight-dispatch* ]]; then
   test_libtorch
 elif [[ "${TEST_CONFIG}" = docs_test ]]; then
   test_docs_test
+elif [[ "${BUILD_ENVIRONMENT}" == *rocm* && -n "$TESTS_TO_INCLUDE" ]]; then
+  install_torchvision
+  test_python
+  test_aten
 elif [[ "${BUILD_ENVIRONMENT}" == *xpu* ]]; then
   install_torchvision
   test_python

.ci/pytorch/win-test-helpers/build_pytorch.bat

@@ -17,22 +17,22 @@ set PATH=C:\Program Files\CMake\bin;C:\Program Files\7-Zip;C:\ProgramData\chocol
 set INSTALLER_DIR=%SCRIPT_HELPERS_DIR%\installation-helpers
 
 call %INSTALLER_DIR%\install_magma.bat
-if errorlevel 1 goto fail
-if not errorlevel 0 goto fail
+if errorlevel 1 exit /b
+if not errorlevel 0 exit /b
 
 call %INSTALLER_DIR%\install_sccache.bat
-if errorlevel 1 goto fail
-if not errorlevel 0 goto fail
+if errorlevel 1 exit /b
+if not errorlevel 0 exit /b
 
 :: Miniconda has been installed as part of the Windows AMI with all the dependencies.
 :: We just need to activate it here
 call %INSTALLER_DIR%\activate_miniconda3.bat
-if errorlevel 1 goto fail
-if not errorlevel 0 goto fail
+if errorlevel 1 exit /b
+if not errorlevel 0 exit /b
 
 call pip install mkl-include==2021.4.0 mkl-devel==2021.4.0
-if errorlevel 1 goto fail
-if not errorlevel 0 goto fail
+if errorlevel 1 exit /b
+if not errorlevel 0 exit /b
 
 :: Override VS env here
 pushd .
@@ -41,8 +41,8 @@ if "%VC_VERSION%" == "" (
 ) else (
     call "C:\Program Files (x86)\Microsoft Visual Studio\%VC_YEAR%\%VC_PRODUCT%\VC\Auxiliary\Build\vcvarsall.bat" x64 -vcvars_ver=%VC_VERSION%
 )
-if errorlevel 1 goto fail
-if not errorlevel 0 goto fail
+if errorlevel 1 exit /b
+if not errorlevel 0 exit /b
 @echo on
 popd
 
@@ -52,12 +52,12 @@ set CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v%CUDA_VERSION%
 
 if x%CUDA_VERSION:.=%==x%CUDA_VERSION% (
     echo CUDA version %CUDA_VERSION% format isn't correct, which doesn't contain '.'
-    goto fail
+    exit /b 1
 )
 rem version transformer, for example 10.1 to 10_1.
 if x%CUDA_VERSION:.=%==x%CUDA_VERSION% (
     echo CUDA version %CUDA_VERSION% format isn't correct, which doesn't contain '.'
-    goto fail
+    exit /b 1
 )
 set VERSION_SUFFIX=%CUDA_VERSION:.=_%
 set CUDA_PATH_V%VERSION_SUFFIX%=%CUDA_PATH%
@@ -101,8 +101,8 @@ if "%USE_CUDA%"=="1" (
   :: CMake requires a single command as CUDA_NVCC_EXECUTABLE, so we push the wrappers
   :: randomtemp.exe and sccache.exe into a batch file which CMake invokes.
   curl -kL https://github.com/peterjc123/randomtemp-rust/releases/download/v0.4/randomtemp.exe --output %TMP_DIR_WIN%\bin\randomtemp.exe
-  if errorlevel 1 goto fail
-  if not errorlevel 0 goto fail
+  if errorlevel 1 exit /b
+  if not errorlevel 0 exit /b
   echo @"%TMP_DIR_WIN%\bin\randomtemp.exe" "%TMP_DIR_WIN%\bin\sccache.exe" "%CUDA_PATH%\bin\nvcc.exe" %%* > "%TMP_DIR%/bin/nvcc.bat"
   cat %TMP_DIR%/bin/nvcc.bat
   set CUDA_NVCC_EXECUTABLE=%TMP_DIR%/bin/nvcc.bat
@@ -114,8 +114,8 @@ if "%USE_CUDA%"=="1" (
 set
 
 python setup.py bdist_wheel
-if errorlevel 1 goto fail
-if not errorlevel 0 goto fail
+if errorlevel 1 exit /b
+if not errorlevel 0 exit /b
 sccache --show-stats
 python -c "import os, glob; os.system('python -mpip install --no-index --no-deps ' + glob.glob('dist/*.whl')[0])"
 (
@@ -135,8 +135,3 @@ python -c "import os, glob; os.system('python -mpip install --no-index --no-deps
 
 sccache --show-stats --stats-format json | jq .stats > sccache-stats-%BUILD_ENVIRONMENT%-%OUR_GITHUB_JOB_ID%.json
 sccache --stop-server
-
-exit /b 0
-
-:fail
-exit /b 1

.ci/pytorch/win-test-helpers/run_python_nn_smoketests.py

@@ -4,7 +4,6 @@ import os
 import subprocess
 import sys
 
-
 COMMON_TESTS = [
     (
         "Checking that torch is available",

.circleci/codegen_validation/normalize_yaml_fragment.py

@@ -5,7 +5,6 @@ import sys
 
 import yaml
 
-
 # Need to import modules that lie on an upward-relative path
 sys.path.append(os.path.join(sys.path[0], ".."))
 

.circleci/scripts/binary_linux_test.sh

@@ -46,18 +46,13 @@ if [[ "\$python_nodot" = *310* ]]; then
   PROTOBUF_PACKAGE="protobuf>=3.19.0"
 fi
 
-if [[ "\$python_nodot" = *39* ]]; then
+if [[ "\$python_nodot" = *39*  ]]; then
   # There's an issue with conda channel priority where it'll randomly pick 1.19 over 1.20
   # we set a lower boundary here just to be safe
   NUMPY_PIN=">=1.20"
 fi
 
-if [[ "\$python_nodot" = *38* ]]; then
-  # sympy 1.12.1 is the last version that supports Python 3.8
-  SYMPY_PIN="==1.12.1"
-else
-  SYMPY_PIN=">=1.13.0"
-fi
+
 
 # Move debug wheels out of the package dir so they don't get installed
 mkdir -p /tmp/debug_final_pkgs
@@ -88,7 +83,7 @@ if [[ "$PACKAGE_TYPE" == conda ]]; then
       "numpy\${NUMPY_PIN}" \
       mkl>=2018 \
       ninja \
-      "sympy\${SYMPY_PIN}" \
+      sympy \
       typing-extensions \
       ${PROTOBUF_PACKAGE}
     if [[ "$DESIRED_CUDA" == 'cpu' ]]; then
@@ -101,21 +96,8 @@ if [[ "$PACKAGE_TYPE" == conda ]]; then
     conda install \${EXTRA_CONDA_FLAGS} -y "\$pkg" --offline
   )
 elif [[ "$PACKAGE_TYPE" != libtorch ]]; then
-  if [[ "\$BUILD_ENVIRONMENT" != *s390x* ]]; then
-    if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-      pkg_no_python="$(ls -1 /final_pkgs/torch_no_python* | sort |tail -1)"
-      pkg_torch="$(ls -1 /final_pkgs/torch-* | sort |tail -1)"
-      # todo: after folder is populated use the pypi_pkg channel instead
-      pip install "\$pkg_no_python" "\$pkg_torch" --index-url "https://download.pytorch.org/whl/\${CHANNEL}/${DESIRED_CUDA}_pypi_pkg"
-      retry pip install -q numpy protobuf typing-extensions
-    else
-      pip install "\$pkg" --index-url "https://download.pytorch.org/whl/\${CHANNEL}/${DESIRED_CUDA}"
-      retry pip install -q numpy protobuf typing-extensions
-    fi
-  else
-    pip install "\$pkg"
-    retry pip install -q numpy protobuf typing-extensions
-  fi
+  pip install "\$pkg" --index-url "https://download.pytorch.org/whl/\${CHANNEL}/${DESIRED_CUDA}"
+  retry pip install -q numpy protobuf typing-extensions
 fi
 if [[ "$PACKAGE_TYPE" == libtorch ]]; then
   pkg="\$(ls /final_pkgs/*-latest.zip)"
@@ -123,18 +105,9 @@ if [[ "$PACKAGE_TYPE" == libtorch ]]; then
   cd /tmp/libtorch
 fi
 
-if [[ "$GPU_ARCH_TYPE" == xpu ]]; then
-  # Workaround for __mkl_tmp_MOD unbound variable issue, refer https://github.com/pytorch/pytorch/issues/130543
-  set +u
-  source /opt/intel/oneapi/pytorch-gpu-dev-0.5/oneapi-vars.sh
-fi
-
 # Test the package
 /builder/check_binary.sh
 
-# Clean temp files
-cd /builder && git clean -ffdx
-
 # =================== The above code will be executed inside Docker container ===================
 EOL
 echo

.circleci/scripts/binary_populate_env.sh

@@ -33,9 +33,9 @@ if [[ -z "$DOCKER_IMAGE" ]]; then
   if [[ "$PACKAGE_TYPE" == conda ]]; then
     export DOCKER_IMAGE="pytorch/conda-cuda"
   elif [[ "$DESIRED_CUDA" == cpu ]]; then
-    export DOCKER_IMAGE="pytorch/manylinux:cpu"
+    export DOCKER_IMAGE="pytorch/manylinux-cpu"
   else
-    export DOCKER_IMAGE="pytorch/manylinux-builder:${DESIRED_CUDA:2}"
+    export DOCKER_IMAGE="pytorch/manylinux-cuda${DESIRED_CUDA:2}"
   fi
 fi
 
@@ -75,9 +75,9 @@ export PYTORCH_BUILD_NUMBER=1
 TRITON_VERSION=$(cat $PYTORCH_ROOT/.ci/docker/triton_version.txt)
 
 # Here PYTORCH_EXTRA_INSTALL_REQUIREMENTS is already set for the all the wheel builds hence append TRITON_CONSTRAINT
-TRITON_CONSTRAINT="platform_system == 'Linux' and platform_machine == 'x86_64' and python_version < '3.13'"
 if [[ "$PACKAGE_TYPE" =~ .*wheel.* &&  -n "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" ]]; then
-  # Only linux Python < 3.13 are supported wheels for triton
+  # Only linux Python < 3.12 are supported wheels for triton
+  TRITON_CONSTRAINT="platform_system == 'Linux' and platform_machine == 'x86_64' and python_version < '3.12'"
   TRITON_REQUIREMENT="triton==${TRITON_VERSION}; ${TRITON_CONSTRAINT}"
   if [[ -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*dev.* ]]; then
       TRITON_SHORTHASH=$(cut -c1-10 $PYTORCH_ROOT/.ci/docker/ci_commit_pins/triton.txt)
@@ -87,11 +87,11 @@ if [[ "$PACKAGE_TYPE" =~ .*wheel.* &&  -n "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:
 fi
 
 # Set triton via PYTORCH_EXTRA_INSTALL_REQUIREMENTS for triton rocm package
-if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*rocm.* && $(uname) == "Linux" ]]; then
-    TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}; ${TRITON_CONSTRAINT}"
+if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*rocm.* && $(uname) == "Linux" && "$DESIRED_PYTHON" != "3.12" ]]; then
+    TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}"
     if [[ -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*dev.* ]]; then
         TRITON_SHORTHASH=$(cut -c1-10 $PYTORCH_ROOT/.ci/docker/ci_commit_pins/triton-rocm.txt)
-        TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}+${TRITON_SHORTHASH}; ${TRITON_CONSTRAINT}"
+        TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}+${TRITON_SHORTHASH}"
     fi
     if [[ -z "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" ]]; then
         export PYTORCH_EXTRA_INSTALL_REQUIREMENTS="${TRITON_REQUIREMENT}"
@@ -100,18 +100,30 @@ if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_B
     fi
 fi
 
-# Set triton via PYTORCH_EXTRA_INSTALL_REQUIREMENTS for triton xpu package
-if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*xpu.* && $(uname) == "Linux" ]]; then
-    TRITON_REQUIREMENT="pytorch-triton-xpu==${TRITON_VERSION}"
-    if [[ -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*dev.* ]]; then
-        TRITON_SHORTHASH=$(cut -c1-10 $PYTORCH_ROOT/.ci/docker/ci_commit_pins/triton-xpu.txt)
-        TRITON_REQUIREMENT="pytorch-triton-xpu==${TRITON_VERSION}+${TRITON_SHORTHASH}"
-    fi
-    if [[ -z "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" ]]; then
-        export PYTORCH_EXTRA_INSTALL_REQUIREMENTS="${TRITON_REQUIREMENT}"
-    else
-        export PYTORCH_EXTRA_INSTALL_REQUIREMENTS="${PYTORCH_EXTRA_INSTALL_REQUIREMENTS} | ${TRITON_REQUIREMENT}"
+JAVA_HOME=
+BUILD_JNI=OFF
+if [[ "$PACKAGE_TYPE" == libtorch ]]; then
+  POSSIBLE_JAVA_HOMES=()
+  POSSIBLE_JAVA_HOMES+=(/usr/local)
+  POSSIBLE_JAVA_HOMES+=(/usr/lib/jvm/java-8-openjdk-amd64)
+  POSSIBLE_JAVA_HOMES+=(/Library/Java/JavaVirtualMachines/*.jdk/Contents/Home)
+  # Add the Windows-specific JNI path
+  POSSIBLE_JAVA_HOMES+=("$PWD/pytorch/.circleci/windows-jni/")
+  for JH in "${POSSIBLE_JAVA_HOMES[@]}" ; do
+    if [[ -e "$JH/include/jni.h" ]] ; then
+      # Skip if we're not on Windows but haven't found a JAVA_HOME
+      if [[ "$JH" == "$PWD/pytorch/.circleci/windows-jni/" && "$OSTYPE" != "msys" ]] ; then
+        break
+      fi
+      echo "Found jni.h under $JH"
+      JAVA_HOME="$JH"
+      BUILD_JNI=ON
+      break
     fi
+  done
+  if [ -z "$JAVA_HOME" ]; then
+    echo "Did not find jni.h"
+  fi
 fi
 
 cat >"$envfile" <<EOL
@@ -124,7 +136,6 @@ export DESIRED_PYTHON="${DESIRED_PYTHON:-}"
 export DESIRED_CUDA="$DESIRED_CUDA"
 export LIBTORCH_VARIANT="${LIBTORCH_VARIANT:-}"
 export BUILD_PYTHONLESS="${BUILD_PYTHONLESS:-}"
-export USE_SPLIT_BUILD="${USE_SPLIT_BUILD:-}"
 if [[ "${OSTYPE}" == "msys" ]]; then
   export LIBTORCH_CONFIG="${LIBTORCH_CONFIG:-}"
   if [[ "${LIBTORCH_CONFIG:-}" == 'debug' ]]; then
@@ -148,6 +159,8 @@ export TORCH_CONDA_BUILD_FOLDER='pytorch-nightly'
 export ANACONDA_USER='pytorch'
 
 export USE_FBGEMM=1
+export JAVA_HOME=$JAVA_HOME
+export BUILD_JNI=$BUILD_JNI
 export PIP_UPLOAD_FOLDER="$PIP_UPLOAD_FOLDER"
 export DOCKER_IMAGE="$DOCKER_IMAGE"
 

.circleci/scripts/binary_upload.sh

@@ -25,15 +25,6 @@ if [[ "${DRY_RUN}" = "disabled" ]]; then
   AWS_S3_CP="aws s3 cp"
 fi
 
-if [[ "${USE_SPLIT_BUILD:-false}" == "true" ]]; then
-  UPLOAD_SUBFOLDER="${UPLOAD_SUBFOLDER}_pypi_pkg"
-fi
-
-# this is special build with all dependencies packaged
-if [[ ${BUILD_NAME} == *-full* ]]; then
-  UPLOAD_SUBFOLDER="${UPLOAD_SUBFOLDER}_full"
-fi
-
 # Sleep 2 minutes between retries for conda upload
 retry () {
   "$@"  || (sleep 5m && "$@") || (sleep 5m && "$@") || (sleep 5m && "$@") || (sleep 5m && "$@")

.circleci/scripts/trigger_azure_pipeline.py

@@ -8,7 +8,6 @@ import time
 
 import requests
 
-
 AZURE_PIPELINE_BASE_URL = "https://aiinfra.visualstudio.com/PyTorch/"
 AZURE_DEVOPS_PAT_BASE64 = os.environ.get("AZURE_DEVOPS_PAT_BASE64_SECRET", "")
 PIPELINE_ID = "911"

.clang-tidy

@@ -36,7 +36,6 @@ hicpp-exception-baseclass,
 hicpp-avoid-goto,
 misc-*,
 -misc-const-correctness,
--misc-include-cleaner,
 -misc-use-anonymous-namespace,
 -misc-unused-parameters,
 -misc-no-recursion,
@@ -61,7 +60,6 @@ readability-simplify-subscript-expr,
 readability-string-compare,
 '
 HeaderFilterRegex: '^(aten/|c10/|torch/).*$'
+AnalyzeTemporaryDtors: false
 WarningsAsErrors: '*'
-CheckOptions:
-  misc-header-include-cycle.IgnoredFilesList: 'format.h;ivalue.h;custom_class.h;Dict.h;List.h'
 ...

.flake8

@@ -2,7 +2,7 @@
 # NOTE: **Mirror any changes** to this file the [tool.ruff] config in pyproject.toml
 # before we can fully move to use ruff
 enable-extensions = G
-select = B,C,E,F,G,P,SIM1,SIM911,T4,W,B9,TOR0,TOR1,TOR2,TOR9
+select = B,C,E,F,G,P,SIM1,T4,W,B9,TOR0,TOR1,TOR2,TOR9
 max-line-length = 120
 # C408 ignored because we like the dict keyword argument syntax
 # E501 is not flexible enough, we're using B950 instead
@@ -54,7 +54,6 @@ per-file-ignores =
     torch/ao/quantization/fx/_decomposed.py: TOR901
     torch/distributed/_functional_collectives.py: TOR901
     torch/distributed/_spmd/data_parallel.py: TOR901
-    torch/distributed/_tensor/_collective_utils.py: TOR901
 optional-ascii-coding = True
 exclude =
     ./.git,

.git-blame-ignore-revs

@@ -40,7 +40,3 @@ e6ec0efaf87703c5f889cfc20b29be455885d58d
 a53cda1ddc15336dc1ff0ce1eff2a49cdc5f882e
 # 2024-01-02 clangformat: fused adam #116583
 9dc68d1aa9e554d09344a10fff69f7b50b2d23a0
-# 2024-06-28 enable UFMT in `torch/storage.py`
-d80939e5e9337e8078f11489afefec59fd42f93b
-# 2024-06-28 enable UFMT in `torch.utils.data`
-7cf0b90e49689d45be91aa539fdf54cf2ea8a9a3

.gitattributes

@@ -4,4 +4,3 @@
 .github/generated-* linguist-generated=true
 .github/scripts/gql_mocks.json linguist-generated=true
 third_party/LICENSES_BUNDLED.txt linguist-generated=true
-tools/build/bazel/requirements.txt linguist-generated=true

.github/ISSUE_TEMPLATE/pt2-bug-report.yml

@@ -8,18 +8,7 @@ body:
       value: >
         #### Before submitting a bug, please make sure the issue hasn't been already addressed by searching through [the
         existing and past issues](https://github.com/pytorch/pytorch/issues)
-        It's likely that your bug will be resolved by checking our FAQ or troubleshooting guide [documentation](https://pytorch.org/docs/main/dynamo/index.html)
-
-        Note: if you're submitting an issue that you generated from a fuzzer. Please do the following:
-
-        - Ensure rtol/atol are at default tolerances
-
-        - Dont compare indices of max/min etc, because that avoids the above requirement
-
-        - If comparing eager and torch.compile at fp16/bf16, you should use fp32 as baseline
-
-        If the above requirements are met, add the label "topic: fuzzer" to your issue.
-
+        It's likely that your bug will be resolved by checking our FAQ or troubleshooting guide [documentation](https://pytorch.org/docs/master/dynamo/index.html)
   - type: textarea
     attributes:
       label: 🐛 Describe the bug
@@ -44,7 +33,7 @@ body:
       label: Minified repro
       description: |
         Please run the minifier on your example and paste the minified code below
-        Learn more here https://pytorch.org/docs/main/torch.compiler_troubleshooting.html
+        Learn more here https://pytorch.org/docs/master/compile/troubleshooting.html
       placeholder: |
         env TORCHDYNAMO_REPRO_AFTER="aot" python your_model.py
         or

.github/actionlint.yaml

@@ -1,12 +1,9 @@
 self-hosted-runner:
   labels:
-    # GitHub hosted x86 Linux runners
     - linux.20_04.4x
     - linux.20_04.16x
-    # Repo-specific LF hosted ARC runners
-    - linux.large.arc
-    # Organization-wide AWS Linux Runners
     - linux.large
+    - linux.large.arc
     - linux.2xlarge
     - linux.4xlarge
     - linux.12xlarge
@@ -16,36 +13,16 @@ self-hosted-runner:
     - linux.8xlarge.nvidia.gpu
     - linux.16xlarge.nvidia.gpu
     - linux.g5.4xlarge.nvidia.gpu
-    # Organization-wide AWS Linux Runners on Linux Foundation account
-    - lf.linux.large
-    - lf.linux.2xlarge
-    - lf.linux.4xlarge
-    - lf.linux.12xlarge
-    - lf.linux.24xlarge
-    - lf.linux.arm64.2xlarge
-    - lf.linux.4xlarge.nvidia.gpu
-    - lf.linux.8xlarge.nvidia.gpu
-    - lf.linux.16xlarge.nvidia.gpu
-    - lf.linux.g5.4xlarge.nvidia.gpu
-    # Repo-specific IBM hosted S390x runner
-    - linux.s390x
-    # Organization wide AWS Windows runners
     - windows.4xlarge.nonephemeral
     - windows.8xlarge.nvidia.gpu
     - windows.8xlarge.nvidia.gpu.nonephemeral
     - windows.g5.4xlarge.nvidia.gpu
-    # Organization-wide AMD hosted MI300 runners
+    - bm-runner
     - linux.rocm.gpu
-    # Repo-specific Apple hosted  runners
-    - macos-m1-ultra
-    - macos-m2-14
-    # Org wise AWS `mac2.metal` runners (2020 Mac mini hardware powered by Apple silicon M1 processors)
     - macos-m1-stable
     - macos-m1-13
-    - macos-m1-14
-    # GitHub-hosted MacOS runners
+    - macos-12-xl
+    - macos-12
+    - macos12.3-m1
     - macos-latest-xlarge
     - macos-13-xlarge
-    - macos-14-xlarge
-    # Organization-wide Intel hosted XPU runners
-    - linux.idc.xpu

.github/actions/diskspace-cleanup/action.yml

@@ -14,14 +14,12 @@ runs:
     - name: Cleans up diskspace
       shell: bash
       run: |
-        set -ex
         diskspace_cutoff=${{ inputs.diskspace-cutoff }}
-        docker_root_dir=$(docker info -f '{{.DockerRootDir}}')
-        diskspace=$(df -H --output=pcent ${docker_root_dir} | sed -n 2p | sed 's/%//' | sed 's/ //')
+        diskspace=$(df -H / --output=pcent | sed -n 2p | sed 's/%//' | sed 's/ //')
         msg="Please file an issue on pytorch/pytorch reporting the faulty runner. Include a link to the runner logs so the runner can be identified"
         if [[ "$diskspace" -ge "$diskspace_cutoff" ]] ; then
             docker system prune -af
-            diskspace_new=$(df -H --output=pcent ${docker_root_dir} | sed -n 2p | sed 's/%//' | sed 's/ //')
+            diskspace_new=$(df -H / --output=pcent | sed -n 2p | sed 's/%//' | sed 's/ //')
             if [[ "$diskspace_new" -gt "$diskspace_cutoff" ]] ; then
                 echo "Error: Available diskspace is less than $diskspace_cutoff percent. Not enough diskspace."
                 echo "$msg"

.github/actions/download-build-artifacts/action.yml

@@ -9,10 +9,6 @@ inputs:
   use-gha:
     description: If set to any value, use GHA to download the artifact. Otherwise use s3.
     required: false
-  s3-bucket:
-    description: S3 bucket to download builds
-    required: false
-    default: "gha-artifacts"
 
 runs:
   using: composite
@@ -22,10 +18,9 @@ runs:
       uses: seemethere/download-artifact-s3@v4
       with:
         name: ${{ inputs.name }}
-        s3-bucket: ${{ inputs.s3-bucket }}
 
     - name: Download PyTorch Build Artifacts from GHA
-      if: ${{ inputs.use-gha }}
+      if: inputs.use-gha
       uses: actions/download-artifact@v3
       with:
         name: ${{ inputs.name }}
@@ -34,10 +29,6 @@ runs:
       shell: bash
       run: unzip -o artifacts.zip
 
-    - name: Remove artifacts.zip
-      shell: bash
-      run: rm artifacts.zip
-
     - name: Output disk space left
       shell: bash
       run: df -H

.github/actions/filter-test-configs/action.yml

@@ -13,13 +13,6 @@ inputs:
     required: true
     type: string
     description: JSON description of what test configs to run.
-  selected-test-configs:
-    required: false
-    type: string
-    description: |
-      A comma-separated list of test configurations from the test matrix to keep,
-      The empty list means we are going to keep every configurations by defaults
-    default: ""
   job-name:
     type: string
     required: false
@@ -47,9 +40,6 @@ outputs:
   ci-no-td:
     description: True if ci-no-td label was on PR or [ci-no-td] in PR body.
     value: ${{ steps.filter.outputs.ci-no-td }}
-  ci-td-distributed:
-    description: True if ci-td-distributed label was on PR or [ci-td-distributed] in PR body.
-    value: ${{ steps.filter.outputs.ci-td-distributed }}
 
 runs:
   using: composite
@@ -66,8 +56,7 @@ runs:
         command: |
           set -eux
           # PyYAML 6.0 doesn't work with MacOS x86 anymore
-          # This must run on Python-3.7 (AmazonLinux2) so can't use request=3.32.2
-          python3 -m pip install requests==2.27.1 pyyaml==6.0.1
+          python3 -m pip install requests==2.26.0 pyyaml==6.0.1
 
     - name: Parse ref
       id: parse-ref
@@ -134,7 +123,6 @@ runs:
           --workflow "${GITHUB_WORKFLOW}" \
           --job-name "${JOB_NAME}" \
           --test-matrix "${{ inputs.test-matrix }}" \
-          --selected-test-configs "${{ inputs.selected-test-configs }}" \
           --pr-number "${PR_NUMBER}" \
           --tag "${TAG}" \
           --event-name "${EVENT_NAME}" \

.github/actions/linux-build/action.yml

@@ -1,226 +0,0 @@
-name: linux-build
-
-inputs:
-  build-environment:
-    required: true
-    description: Top-level label for what's being built/tested.
-  docker-image-name:
-    required: true
-    description: Name of the base docker image to build with.
-  build-generates-artifacts:
-    required: false
-    default: "true"
-    description: If set, upload generated build artifacts.
-  build-with-debug:
-    required: false
-    default: "false"
-    description: If set, build in debug mode.
-  sync-tag:
-    required: false
-    default: ""
-    description: |
-      If this is set, our linter will use this to make sure that every other
-      job with the same `sync-tag` is identical.
-  cuda-arch-list:
-    required: false
-    default: "5.2"
-    description: Runner label to select worker type
-  runner:
-    required: false
-    default: "linux.2xlarge"
-    description: |
-      List of CUDA architectures CI build should target.
-  test-matrix:
-    required: false
-    type: string
-    description: |
-      An option JSON description of what test configs to run later on. This
-      is moved here from the Linux test workflow so that we can apply filter
-      logic using test-config labels earlier and skip unnecessary builds
-  s3-bucket:
-    description: S3 bucket to download artifact
-    required: false
-    default: "gha-artifacts"
-  aws-role-to-assume:
-    description: role to assume for downloading artifacts
-    required: false
-    default: ""
-  GITHUB_TOKEN:
-    description: GitHub token
-    required: true
-  HUGGING_FACE_HUB_TOKEN:
-    description: Hugging Face Hub token
-    required: false
-    default: ""
-  use_split_build:
-    description: |
-      [Experimental] Build a libtorch only wheel and build pytorch such that
-      are built from the libtorch wheel.
-    required: false
-    type: boolean
-    default: false
-outputs:
-  docker-image:
-    value: ${{ steps.calculate-docker-image.outputs.docker-image }}
-    description: The docker image containing the built PyTorch.
-  test-matrix:
-    value: ${{ steps.filter.outputs.test-matrix }}
-    description: An optional JSON description of what test configs to run later on.
-
-runs:
-  using: composite
-  steps:
-    - name: Setup Linux
-      uses: ./.github/actions/setup-linux
-
-    - name: configure aws credentials
-      uses: aws-actions/configure-aws-credentials@v3
-      if: ${{ inputs.aws-role-to-assume != '' }}
-      with:
-        role-to-assume: ${{ inputs.aws-role-to-assume }}
-        role-session-name: gha-linux-build
-        role-duration-seconds: 10800
-        aws-region: us-east-1
-
-    - name: Calculate docker image
-      id: calculate-docker-image
-      uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
-      with:
-        docker-image-name: ${{ inputs.docker-image-name }}
-
-    - name: Use following to pull public copy of the image
-      id: print-ghcr-mirror
-      env:
-        ECR_DOCKER_IMAGE: ${{ steps.calculate-docker-image.outputs.docker-image }}
-      shell: bash
-      run: |
-        tag=${ECR_DOCKER_IMAGE##*/}
-        echo "docker pull ghcr.io/pytorch/ci-image:${tag/:/-}"
-
-    - name: Pull docker image
-      uses: pytorch/test-infra/.github/actions/pull-docker-image@main
-      with:
-        docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
-
-    - name: Parse ref
-      id: parse-ref
-      shell: bash
-      run: .github/scripts/parse_ref.py
-
-    - name: Get workflow job id
-      id: get-job-id
-      uses: ./.github/actions/get-workflow-job-id
-      if: always()
-      with:
-        github-token: ${{ inputs.GITHUB_TOKEN }}
-
-    # Apply the filter logic to the build step too if the test-config label is already there
-    - name: Select all requested test configurations (if the test matrix is available)
-      id: filter
-      uses: ./.github/actions/filter-test-configs
-      with:
-        github-token: ${{ inputs.GITHUB_TOKEN }}
-        test-matrix: ${{ inputs.test-matrix }}
-        job-name: ${{ steps.get-job-id.outputs.job-name }}
-
-    - name: Download pytest cache
-      uses: ./.github/actions/pytest-cache-download
-      continue-on-error: true
-      with:
-        cache_dir: .pytest_cache
-        job_identifier: ${{ github.workflow }}_${{ inputs.build-environment }}
-        s3_bucket: ${{ inputs.s3-bucket }}
-
-    - name: Build
-      if: steps.filter.outputs.is-test-matrix-empty == 'False' || inputs.test-matrix == ''
-      id: build
-      env:
-        BUILD_ENVIRONMENT: ${{ inputs.build-environment }}
-        BRANCH: ${{ steps.parse-ref.outputs.branch }}
-        # TODO duplicated
-        AWS_DEFAULT_REGION: us-east-1
-        PR_NUMBER: ${{ github.event.pull_request.number }}
-        SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
-        SCCACHE_BUCKET: ossci-compiler-cache-circleci-v2
-        SCCACHE_S3_KEY_PREFIX: ${{ github.workflow }}
-        XLA_CLANG_CACHE_S3_BUCKET_NAME: ossci-compiler-clang-cache-circleci-xla
-        PR_LABELS: ${{ toJson(github.event.pull_request.labels.*.name) }}
-        TORCH_CUDA_ARCH_LIST: ${{ inputs.cuda-arch-list }}
-        DOCKER_IMAGE: ${{ steps.calculate-docker-image.outputs.docker-image }}
-        XLA_CUDA: ${{ contains(inputs.build-environment, 'xla') && '0' || '' }}
-        DEBUG: ${{ inputs.build-with-debug == 'true' && '1' || '0' }}
-        OUR_GITHUB_JOB_ID: ${{ steps.get-job-id.outputs.job-id }}
-        HUGGING_FACE_HUB_TOKEN: ${{ inputs.HUGGING_FACE_HUB_TOKEN }}
-        USE_SPLIT_BUILD: ${{ inputs.use_split_build }}
-      shell: bash
-      run: |
-        # detached container should get cleaned up by teardown_ec2_linux
-        container_name=$(docker run \
-          -e BUILD_ENVIRONMENT \
-          -e MAX_JOBS="$(nproc --ignore=2)" \
-          -e AWS_DEFAULT_REGION \
-          -e PR_NUMBER \
-          -e SHA1 \
-          -e BRANCH \
-          -e SCCACHE_BUCKET \
-          -e SCCACHE_S3_KEY_PREFIX \
-          -e XLA_CUDA \
-          -e XLA_CLANG_CACHE_S3_BUCKET_NAME \
-          -e SKIP_SCCACHE_INITIALIZATION=1 \
-          -e TORCH_CUDA_ARCH_LIST \
-          -e PR_LABELS \
-          -e OUR_GITHUB_JOB_ID \
-          -e HUGGING_FACE_HUB_TOKEN \
-          -e USE_SPLIT_BUILD \
-          --env-file="/tmp/github_env_${GITHUB_RUN_ID}" \
-          --security-opt seccomp=unconfined \
-          --cap-add=SYS_PTRACE \
-          --tty \
-          --detach \
-          --user jenkins \
-          -v "${GITHUB_WORKSPACE}:/var/lib/jenkins/workspace" \
-          -w /var/lib/jenkins/workspace \
-          "${DOCKER_IMAGE}"
-        )
-        docker exec -t "${container_name}" sh -c '.ci/pytorch/build.sh'
-
-    - name: Archive artifacts into zip
-      if: inputs.build-generates-artifacts == 'true' && steps.build.outcome != 'skipped'
-      shell: bash
-      run: |
-        zip -1 -r artifacts.zip dist/ build/custom_test_artifacts build/lib build/bin .additional_ci_files
-
-    - name: Store PyTorch Build Artifacts on S3
-      uses: seemethere/upload-artifact-s3@v5
-      if: inputs.build-generates-artifacts == 'true' && steps.build.outcome != 'skipped' && inputs.use_split_build != 'true'
-      with:
-        name: ${{ inputs.build-environment }}
-        retention-days: 14
-        if-no-files-found: error
-        path: artifacts.zip
-        s3-bucket: ${{ inputs.s3-bucket }}
-
-    - name: Store PyTorch Build Artifacts on S3 for split build
-      uses: seemethere/upload-artifact-s3@v5
-      if: inputs.build-generates-artifacts == 'true' && steps.build.outcome != 'skipped' && inputs.use_split_build == 'true'
-      with:
-        name: ${{ inputs.build-environment }}-experimental-split-build
-        retention-days: 14
-        if-no-files-found: error
-        path: artifacts.zip
-        s3-bucket: ${{ inputs.s3-bucket }}
-
-    - name: Upload sccache stats
-      if: steps.build.outcome != 'skipped'
-      uses: seemethere/upload-artifact-s3@v5
-      with:
-        s3-prefix: |
-          ${{ github.repository }}/${{ github.run_id }}/${{ github.run_attempt }}/artifact
-        retention-days: 365
-        if-no-files-found: warn
-        path: sccache-stats-*.json
-        s3-bucket: ${{ inputs.s3-bucket }}
-
-    - name: Teardown Linux
-      uses: pytorch/test-infra/.github/actions/teardown-linux@main
-      if: always()

.github/actions/linux-test/action.yml

@@ -1,384 +0,0 @@
-name: linux-test
-
-inputs:
-  build-environment:
-    required: true
-    type: string
-    description: Top-level label for what's being built/tested.
-  test-matrix:
-    required: true
-    type: string
-    description: JSON description of what test configs to run.
-  docker-image:
-    required: true
-    type: string
-    description: Docker image to run in.
-  sync-tag:
-    required: false
-    type: string
-    default: ""
-    description: |
-      If this is set, our linter will use this to make sure that every other
-      job with the same `sync-tag` is identical.
-  use-gha:
-    required: false
-    type: string
-    default: ""
-    description: If set to any value, upload to GHA. Otherwise upload to S3.
-  dashboard-tag:
-    required: false
-    type: string
-    default: ""
-  s3-bucket:
-    description: S3 bucket to download artifact
-    required: false
-    type: string
-    default: "gha-artifacts"
-  aws-role-to-assume:
-    description: role to assume for downloading artifacts
-    required: false
-    type: string
-    default: ""
-  HUGGING_FACE_HUB_TOKEN:
-    description: |
-      HF Auth token to avoid rate limits when downloading models or datasets from hub
-    required: false
-    default: ""
-  GITHUB_TOKEN:
-    description: GitHub token
-    required: true
-
-#env:
-#  GIT_DEFAULT_BRANCH: ${{ inputs.default_branch }}
-
-runs:
-  using: composite
-  steps:
-    - name: Setup Linux
-      uses: ./.github/actions/setup-linux
-
-    - name: configure aws credentials
-      if : ${{ inputs.aws-role-to-assume != '' }}
-      uses: aws-actions/configure-aws-credentials@v3
-      with:
-        role-to-assume: ${{ inputs.aws-role-to-assume }}
-        role-session-name: gha-linux-test
-        aws-region: us-east-1
-
-    - name: Calculate docker image
-      id: calculate-docker-image
-      uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
-      with:
-        docker-image-name: ${{ inputs.docker-image }}
-
-    - name: Use following to pull public copy of the image
-      id: print-ghcr-mirror
-      env:
-        ECR_DOCKER_IMAGE: ${{ steps.calculate-docker-image.outputs.docker-image }}
-      shell: bash
-      run: |
-        tag=${ECR_DOCKER_IMAGE##*/}
-        echo "docker pull ghcr.io/pytorch/ci-image:${tag/:/-}"
-
-    - name: Pull docker image
-      uses: pytorch/test-infra/.github/actions/pull-docker-image@main
-      with:
-        docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
-
-    - name: Check if in a ARC runner
-      shell: bash
-      id: check_arc_runner
-      run: echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)" >> "$GITHUB_OUTPUT"
-
-    - name: Install nvidia driver, nvidia-docker runtime, set GPU_FLAG
-      id: install-nvidia-driver
-      uses: pytorch/test-infra/.github/actions/setup-nvidia@main
-      if: ${{ contains(inputs.build-environment, 'cuda') && !contains(matrix.config, 'nogpu') && steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
-
-    - name: Lock NVIDIA A100 40GB Frequency
-      shell: bash
-      run: |
-        sudo nvidia-smi -pm 1
-        sudo nvidia-smi -ac 1215,1410
-        nvidia-smi
-      if: contains(matrix.runner, 'a100')
-
-    - name: Start monitoring script
-      id: monitor-script
-      shell: bash
-      continue-on-error: true
-      run: |
-        python3 -m pip install psutil==5.9.1 nvidia-ml-py==11.525.84
-        python3 -m tools.stats.monitor > usage_log.txt 2>&1 &
-        echo "monitor-script-pid=${!}" >> "${GITHUB_OUTPUT}"
-
-    - name: Download build artifacts
-      uses: ./.github/actions/download-build-artifacts
-      with:
-        name: ${{ inputs.build-environment }}
-        s3-bucket: ${{ inputs.s3-bucket }}
-
-    - name: Download TD artifacts
-      continue-on-error: true
-      uses: ./.github/actions/download-td-artifacts
-
-    - name: Parse ref
-      id: parse-ref
-      shell: bash
-      run: .github/scripts/parse_ref.py
-
-    - name: Get workflow job id
-      id: get-job-id
-      uses: ./.github/actions/get-workflow-job-id
-      if: always()
-      with:
-        github-token: ${{ inputs.GITHUB_TOKEN }}
-
-    - name: Check for keep-going label and re-enabled test issues
-      # This uses the filter-test-configs action because it conviniently
-      # checks for labels and re-enabled test issues.  It does not actually do
-      # any filtering.  All filtering is done in the build step.
-      id: keep-going
-      uses: ./.github/actions/filter-test-configs
-      with:
-        github-token: ${{ inputs.GITHUB_TOKEN }}
-        test-matrix: ${{ inputs.test-matrix }}
-        job-name: ${{ steps.get-job-id.outputs.job-name }}
-
-    - name: Test
-      id: test
-      env:
-        BUILD_ENVIRONMENT: ${{ inputs.build-environment }}
-        PR_NUMBER: ${{ github.event.pull_request.number }}
-        GITHUB_REPOSITORY: ${{ github.repository }}
-        GITHUB_WORKFLOW: ${{ github.workflow }}
-        GITHUB_JOB: ${{ github.job }}
-        GITHUB_RUN_ID: ${{ github.run_id }}
-        GITHUB_RUN_NUMBER: ${{ github.run_number }}
-        GITHUB_RUN_ATTEMPT: ${{ github.run_attempt }}
-        JOB_ID: ${{ steps.get-job-id.outputs.job-id }}
-        JOB_NAME: ${{ steps.get-job-id.outputs.job-name }}
-        BRANCH: ${{ steps.parse-ref.outputs.branch }}
-        SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
-        BASE_SHA: ${{ github.event.pull_request.base.sha || github.sha }}
-        TEST_CONFIG: ${{ matrix.config }}
-        SHARD_NUMBER: ${{ matrix.shard }}
-        NUM_TEST_SHARDS: ${{ matrix.num_shards }}
-        REENABLED_ISSUES: ${{ steps.keep-going.outputs.reenabled-issues }}
-        CONTINUE_THROUGH_ERROR: ${{ steps.keep-going.outputs.keep-going }}
-        VERBOSE_TEST_LOGS: ${{ steps.keep-going.outputs.ci-verbose-test-logs }}
-        NO_TEST_TIMEOUT: ${{ steps.keep-going.outputs.ci-no-test-timeout }}
-        NO_TD: ${{ steps.keep-going.outputs.ci-no-td }}
-        TD_DISTRIBUTED: ${{ steps.keep-going.outputs.ci-td-distributed }}
-        SCCACHE_BUCKET: ossci-compiler-cache-circleci-v2
-        SCCACHE_S3_KEY_PREFIX: ${{ github.workflow }}
-        SHM_SIZE: ${{ contains(inputs.build-environment, 'cuda') && '2g' || '1g' }}
-        DOCKER_IMAGE: ${{ inputs.docker-image }}
-        XLA_CUDA: ${{ contains(inputs.build-environment, 'xla') && '0' || '' }}
-        XLA_CLANG_CACHE_S3_BUCKET_NAME: ossci-compiler-clang-cache-circleci-xla
-        PYTORCH_TEST_CUDA_MEM_LEAK_CHECK: ${{ matrix.mem_leak_check && '1' || '0' }}
-        PYTORCH_TEST_RERUN_DISABLED_TESTS: ${{ matrix.rerun_disabled_tests && '1' || '0' }}
-        DASHBOARD_TAG: ${{ inputs.dashboard-tag }}
-        HUGGING_FACE_HUB_TOKEN: ${{ inputs.HUGGING_FACE_HUB_TOKEN }}
-      shell: bash
-      run: |
-        set -x
-
-        if [[ $TEST_CONFIG == 'multigpu' ]]; then
-          TEST_COMMAND=.ci/pytorch/multigpu-test.sh
-        elif [[ $BUILD_ENVIRONMENT == *onnx* ]]; then
-          TEST_COMMAND=.ci/onnx/test.sh
-        else
-          TEST_COMMAND=.ci/pytorch/test.sh
-        fi
-
-        # detached container should get cleaned up by teardown_ec2_linux
-        # TODO: Stop building test binaries as part of the build phase
-        # Used for GPU_FLAG since that doesn't play nice
-        # shellcheck disable=SC2086,SC2090
-        container_name=$(docker run \
-          ${GPU_FLAG:-} \
-          -e BUILD_ENVIRONMENT \
-          -e PR_NUMBER \
-          -e GITHUB_ACTIONS \
-          -e GITHUB_REPOSITORY \
-          -e GITHUB_WORKFLOW \
-          -e GITHUB_JOB \
-          -e GITHUB_RUN_ID \
-          -e GITHUB_RUN_NUMBER \
-          -e GITHUB_RUN_ATTEMPT \
-          -e JOB_ID \
-          -e JOB_NAME \
-          -e BASE_SHA \
-          -e BRANCH \
-          -e SHA1 \
-          -e AWS_DEFAULT_REGION \
-          -e IN_WHEEL_TEST \
-          -e SHARD_NUMBER \
-          -e TEST_CONFIG \
-          -e NUM_TEST_SHARDS \
-          -e REENABLED_ISSUES \
-          -e CONTINUE_THROUGH_ERROR \
-          -e VERBOSE_TEST_LOGS \
-          -e NO_TEST_TIMEOUT \
-          -e NO_TD \
-          -e TD_DISTRIBUTED \
-          -e PR_LABELS \
-          -e MAX_JOBS="$(nproc --ignore=2)" \
-          -e SCCACHE_BUCKET \
-          -e SCCACHE_S3_KEY_PREFIX \
-          -e XLA_CUDA \
-          -e XLA_CLANG_CACHE_S3_BUCKET_NAME \
-          -e PYTORCH_TEST_CUDA_MEM_LEAK_CHECK \
-          -e PYTORCH_TEST_RERUN_DISABLED_TESTS \
-          -e SKIP_SCCACHE_INITIALIZATION=1 \
-          -e HUGGING_FACE_HUB_TOKEN \
-          -e DASHBOARD_TAG \
-          --env-file="/tmp/github_env_${GITHUB_RUN_ID}" \
-          --security-opt seccomp=unconfined \
-          --cap-add=SYS_PTRACE \
-          --ipc=host \
-          --shm-size="${SHM_SIZE}" \
-          --tty \
-          --detach \
-          --name="${container_name}" \
-          --user jenkins \
-          -v "${GITHUB_WORKSPACE}:/var/lib/jenkins/workspace" \
-          -w /var/lib/jenkins/workspace \
-          "${DOCKER_IMAGE}"
-        )
-        # Propagate download.pytorch.org IP to container
-        grep download.pytorch.org /etc/hosts | docker exec -i "${container_name}" sudo bash -c "/bin/cat >> /etc/hosts"
-        echo "DOCKER_CONTAINER_ID=${container_name}" >> "${GITHUB_ENV}"
-        docker exec -t "${container_name}" sh -c "pip install $(echo dist/*.whl)[opt-einsum] && ${TEST_COMMAND}"
-
-    - name: Upload pytest cache if tests failed
-      uses: ./.github/actions/pytest-cache-upload
-      continue-on-error: true
-      if: failure() && steps.test.conclusion && steps.test.conclusion == 'failure'
-      with:
-        cache_dir: .pytest_cache
-        shard: ${{ matrix.shard }}
-        sha: ${{ github.event.pull_request.head.sha || github.sha }}
-        test_config: ${{ matrix.config }}
-        job_identifier: ${{ github.workflow }}_${{ inputs.build-environment }}
-
-    - name: Print remaining test logs
-      shell: bash
-      if: always() && steps.test.conclusion
-      run: |
-        cat test/**/*_toprint.log || true
-
-    - name: Stop monitoring script
-      if: always() && steps.monitor-script.outputs.monitor-script-pid
-      shell: bash
-      continue-on-error: true
-      env:
-        MONITOR_SCRIPT_PID: ${{ steps.monitor-script.outputs.monitor-script-pid }}
-      run: |
-        kill "$MONITOR_SCRIPT_PID"
-
-    - name: Upload test artifacts
-      uses: ./.github/actions/upload-test-artifacts
-      if: always() && steps.test.conclusion && steps.test.conclusion != 'skipped'
-      with:
-        file-suffix: ${{ github.job }}-${{ matrix.config }}-${{ matrix.shard }}-${{ matrix.num_shards }}-${{ matrix.runner }}_${{ steps.get-job-id.outputs.job-id }}
-        use-gha: ${{ inputs.use-gha }}
-        s3-bucket: ${{ inputs.s3-bucket }}
-
-    - name: Collect backtraces from coredumps (if any)
-      if: always()
-      shell: bash
-      run: |
-        # shellcheck disable=SC2156
-        find . -iname "core.[1-9]*" -exec docker exec "${DOCKER_CONTAINER_ID}" sh -c "gdb python {} -ex 'bt' -ex 'q'" \;
-
-    - name: Store Core dumps on S3
-      uses: seemethere/upload-artifact-s3@v5
-      if: failure()
-      with:
-        name: coredumps-${{ matrix.config }}-${{ matrix.shard }}-${{ matrix.num_shards }}-${{ matrix.runner }}
-        retention-days: 14
-        if-no-files-found: ignore
-        path: ./**/core.[1-9]*
-
-    - name: Teardown Linux
-      uses: pytorch/test-infra/.github/actions/teardown-linux@main
-      if: always()
-
-    # NB: We are currently having an intermittent GPU-related issue on G5 runners with
-    # A10G GPU. Once this happens, trying to reset the GPU as done in setup-nvidia does
-    # not seem to help. Here are some symptoms:
-    #   * Calling nvidia-smi timeouts after 60 second
-    #   * Fail to run nvidia-smi with an unable to determine the device handle for GPU
-    #     unknown error
-    #   * Test fails with a missing CUDA GPU error when initializing CUDA in PyTorch
-    #   * Run docker --gpus all fails with error response from daemon
-    #
-    # As both the root cause and recovery path are unclear, let's take the runner out of
-    # service so that it doesn't get any more jobs
-    - name: Check NVIDIA driver installation step
-      if: failure() && steps.install-nvidia-driver.outcome && steps.install-nvidia-driver.outcome != 'skipped'
-      shell: bash
-      env:
-        RUNNER_WORKSPACE: ${{ runner.workspace }}
-      run: |
-        set +e
-        set -x
-
-        nvidia-smi
-        # NB: Surprisingly, nvidia-smi command returns successfully with return code 0 even in
-        # the case where the driver has already crashed as it still can get the driver version
-        # and some basic information like the bus ID.  However, the rest of the information
-        # would be missing (ERR!), for example:
-        #
-        # +-----------------------------------------------------------------------------+
-        # | NVIDIA-SMI 525.89.02    Driver Version: 525.89.02    CUDA Version: 12.0     |
-        # |-------------------------------+----------------------+----------------------+
-        # | GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
-        # | Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
-        # |                               |                      |               MIG M. |
-        # |===============================+======================+======================|
-        # |   0  ERR!                Off  | 00000000:00:1E.0 Off |                 ERR! |
-        # |ERR!  ERR! ERR!    ERR! / ERR! |   4184MiB / 23028MiB |    ERR!      Default |
-        # |                               |                      |                 ERR! |
-        # +-------------------------------+----------------------+----------------------+
-        #
-        # +-----------------------------------------------------------------------------+
-        # | Processes:                                                                  |
-        # |  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
-        # |        ID   ID                                                   Usage      |
-        # |=============================================================================|
-        # +-----------------------------------------------------------------------------+
-        #
-        # This should be reported as a failure instead as it will guarantee to fail when
-        # Docker tries to run with --gpus all
-        #
-        # So, the correct check here is to query one of the missing piece of info like
-        # GPU name, so that the command can fail accordingly
-        nvidia-smi --query-gpu=gpu_name --format=csv,noheader --id=0
-        NVIDIA_SMI_STATUS=$?
-
-        # These are acceptable return code from nvidia-smi as copied from setup-nvidia GitHub action
-        if [ "$NVIDIA_SMI_STATUS" -ne 0 ] && [ "$NVIDIA_SMI_STATUS" -ne 14 ]; then
-          echo "NVIDIA driver installation has failed, shutting down the runner..."
-          .github/scripts/stop_runner_service.sh
-        fi
-
-        # For runner with multiple GPUs, we also want to confirm that the number of GPUs are the
-        # power of 2, i.e. 1, 2, 4, or 8. This is to avoid flaky test issue when one GPU fails
-        # https://github.com/pytorch/test-infra/issues/4000
-        GPU_COUNT=$(nvidia-smi --list-gpus | wc -l)
-        NVIDIA_SMI_STATUS=$?
-
-        # These are acceptable return code from nvidia-smi as copied from setup-nvidia GitHub action
-        if [ "$NVIDIA_SMI_STATUS" -ne 0 ] && [ "$NVIDIA_SMI_STATUS" -ne 14 ]; then
-          echo "NVIDIA driver installation has failed, shutting down the runner..."
-          .github/scripts/stop_runner_service.sh
-        fi
-
-        # Check the GPU count to be a power of 2
-        if [ "$GPU_COUNT" -le 8 ] && [ "$GPU_COUNT" -ne 1 ] && [ "$GPU_COUNT" -ne 2 ] && [ "$GPU_COUNT" -ne 4 ] && [ "$GPU_COUNT" -ne 8 ]; then
-          echo "NVIDIA driver detects $GPU_COUNT GPUs. The runner has a broken GPU, shutting it down..."
-          .github/scripts/stop_runner_service.sh
-        fi

.github/actions/pytest-cache-download/action.yml

@@ -9,10 +9,6 @@ inputs:
   job_identifier:
     description: Text that uniquely identifies a given job type within a workflow. All shards of a job should share the same job identifier.
     required: true
-  s3_bucket:
-    description: S3 bucket to download PyTest cache
-    required: false
-    default: "gha-artifacts"
 
 runs:
   using: composite
@@ -34,7 +30,6 @@ runs:
         CACHE_DIR: ${{ inputs.cache_dir }}
         JOB_IDENTIFIER: ${{ inputs.job_identifier }}
         REPO: ${{ github.repository }}
-        BUCKET: ${{ inputs.s3_bucket }}
       run: |
         python3 .github/scripts/pytest_cache.py \
           --download \
@@ -43,4 +38,3 @@ runs:
           --job_identifier $JOB_IDENTIFIER \
           --temp_dir $RUNNER_TEMP \
           --repo $REPO \
-          --bucket $BUCKET \

.github/actions/setup-linux/action.yml

@@ -15,12 +15,10 @@ runs:
           category=$1
           # If it is GCP runner (runner name contains gcp), do not run this
           runner_name_str=${{ runner.name }}
-          if [[ -f /.inarc ]]; then
-            echo "ARC Runner, no info on ec2 metadata"
-          elif [[ $runner_name_str == *"gcp"* ]]; then
-            echo "Runner is from Google Cloud Platform, No info on ec2 metadata"
-          else
+          if [[ $runner_name_str != *"gcp"* ]]; then
             curl -fsSL "http://169.254.169.254/latest/meta-data/${category}"
+          else
+            echo "Runner is from Google Cloud Platform, No info on ec2 metadata"
           fi
         }
         echo "ami-id: $(get_ec2_metadata ami-id)"
@@ -28,14 +26,8 @@ runs:
         echo "instance-type: $(get_ec2_metadata instance-type)"
         echo "system info $(uname -a)"
 
-    - name: Check if in a ARC runner
-      shell: bash
-      id: check_arc_runner
-      run: echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)"  >> $GITHUB_OUTPUT
-
     - name: Start docker if docker deamon is not running
       shell: bash
-      if: ${{ steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
       run: |
         if systemctl is-active --quiet docker; then
             echo "Docker daemon is running...";
@@ -66,7 +58,6 @@ runs:
         env | grep '^CI' >> "/tmp/github_env_${GITHUB_RUN_ID}"
 
     - name: Kill any existing containers, clean up images
-      if: ${{ steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
       shell: bash
       run: |
         # ignore expansion of "docker ps -q" since it could be empty
@@ -105,28 +96,3 @@ runs:
 
         echo "${RESOLVED_IP} ${PT_DOMAIN}" | sudo tee -a /etc/hosts
         cat /etc/hosts
-
-    - name: Check that the docker daemon is running
-      shell: bash
-      continue-on-error: true
-      if: ${{ steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'true' }}
-      run: |
-        set +x
-
-        max_attempts=30
-        delay=10
-        attempt=1
-
-        for attempt in $(seq 1 $max_attempts); do
-          echo "Attempt $attempt of $max_attempts: Checking if Docker daemon is running..."
-          if docker info > /dev/null 2>&1; then
-            echo "Docker is running. Proceeding with the next steps"
-            exit 0
-          else
-            echo "Docker is not running yet."
-            echo "Retrying in $delay seconds..."
-            sleep $delay
-          fi
-        done
-        echo "Reached maximum attempts to connect to Docker. Exiting."
-        exit 1

.github/actions/test-pytorch-binary/action.yml

@@ -26,7 +26,6 @@ runs:
           -e PYTORCH_FINAL_PACKAGE_DIR \
           -e PYTORCH_ROOT \
           -e SKIP_ALL_TESTS \
-          -e USE_SPLIT_BUILD \
           --tty \
           --detach \
           -v "${GITHUB_WORKSPACE}/pytorch:/pytorch" \
@@ -36,8 +35,7 @@ runs:
           "${DOCKER_IMAGE}"
         )
 
-        echo "CONTAINER_NAME=${container_name}" >> "$GITHUB_ENV"
-        if [[ "${GPU_ARCH_TYPE}" != "rocm" && "${BUILD_ENVIRONMENT}" != "linux-aarch64-binary-manywheel" && "${BUILD_ENVIRONMENT}" != "linux-s390x-binary-manywheel" && "${GPU_ARCH_TYPE}" != "xpu" ]]; then
+        if [[ "${GPU_ARCH_TYPE}" != "rocm" && "${BUILD_ENVIRONMENT}" != "linux-aarch64-binary-manywheel" ]]; then
           # Propagate download.pytorch.org IP to container. This is only needed on Linux non aarch64 runner
           grep download.pytorch.org /etc/hosts | docker exec -i "${container_name}" bash -c "/bin/cat >> /etc/hosts"
         fi
@@ -46,11 +44,3 @@ runs:
         # Generate test script
         docker exec -t -w "${PYTORCH_ROOT}" -e OUTPUT_SCRIPT="/run.sh" "${container_name}" bash -c "bash .circleci/scripts/binary_linux_test.sh"
         docker exec -t "${container_name}" bash -c "source ${BINARY_ENV_FILE} && bash -x /run.sh"
-
-    - name: Cleanup docker
-      if: always() && (env.BUILD_ENVIRONMENT == 'linux-s390x-binary-manywheel' || env.GPU_ARCH_TYPE == 'xpu')
-      shell: bash
-      run: |
-        # on s390x or xpu stop the container for clean worker stop
-        # shellcheck disable=SC2046
-        docker stop "${{ env.CONTAINER_NAME }}" || true

.github/actions/upload-test-artifacts/action.yml

@@ -11,10 +11,6 @@ inputs:
       Suffix to add to the filename of the artifacts. This should include the
       workflow job id, see [Job id in artifacts].
     required: true
-  s3-bucket:
-    description: S3 bucket to download builds
-    required: false
-    default: "gha-artifacts"
 
 runs:
   using: composite
@@ -46,7 +42,7 @@ runs:
       env:
         FILE_SUFFIX: ${{ inputs.file-suffix }}
       run: |
-        # Remove any previous usage logs if they exist
+        # Remove any previous test reports if they exist
         rm -f logs-*.zip
         # this workflow is also run in bazel build test, but we dont generate usage reports for it
         # so check to see if the file exists first
@@ -57,18 +53,6 @@ runs:
             zip -r "logs-${FILE_SUFFIX}.zip" test -i '*.log'
         fi
 
-    - name: Zip debugging artifacts for upload
-      if: runner.os != 'Windows' && !inputs.use-gha
-      shell: bash
-      env:
-        FILE_SUFFIX: ${{ inputs.file-suffix }}
-      run: |
-        # Remove any previous debugging artifacts if they exist
-        rm -f debug-*.zip
-        if [ -d 'test/debug' ]; then
-          zip -r "debug-${FILE_SUFFIX}.zip" test/debug
-        fi
-
     # Windows zip
     - name: Zip JSONs for upload
       if: runner.os == 'Windows' && !inputs.use-gha
@@ -103,7 +87,6 @@ runs:
       uses: seemethere/upload-artifact-s3@v5
       if: ${{ !inputs.use-gha }}
       with:
-        s3-bucket: ${{ inputs.s3-bucket }}
         s3-prefix: |
           ${{ github.repository }}/${{ github.run_id }}/${{ github.run_attempt }}/artifact
         retention-days: 14
@@ -114,7 +97,6 @@ runs:
       uses: seemethere/upload-artifact-s3@v5
       if: ${{ !inputs.use-gha }}
       with:
-        s3-bucket: ${{ inputs.s3-bucket }}
         s3-prefix: |
           ${{ github.repository }}/${{ github.run_id }}/${{ github.run_attempt }}/artifact
         retention-days: 14
@@ -126,25 +108,12 @@ runs:
       if: ${{ !inputs.use-gha }}
       continue-on-error: true
       with:
-        s3-bucket: ${{ inputs.s3-bucket }}
         s3-prefix: |
           ${{ github.repository }}/${{ github.run_id }}/${{ github.run_attempt }}/artifact
         retention-days: 14
         if-no-files-found: ignore
         path: logs-*.zip
 
-    - name: Store Debug Artifacts on S3
-      uses: seemethere/upload-artifact-s3@v5
-      if: ${{ !inputs.use-gha }}
-      continue-on-error: true
-      with:
-        s3-bucket: ${{ inputs.s3-bucket }}
-        s3-prefix: |
-          ${{ github.repository }}/${{ github.run_id }}/${{ github.run_attempt }}/artifact
-        retention-days: 14
-        if-no-files-found: ignore
-        path: debug-*.zip
-
     # GHA upload
     - name: Store Test Downloaded JSONs on Github
       uses: actions/upload-artifact@v3

.github/ci_commit_pins/audio.txt

@@ -1 +1 @@
-69b2a0adc2ec03ab99990d7e8be3d4510438c148
+87aeb554d3e2f7855b7abe5120c282f59648ed7a

.github/ci_commit_pins/torchbench.txt

@@ -1 +1 @@
-23512dbebd44a11eb84afbf53c3c071dd105297e
+d6015d42d9a1834bc7595c4bd6852562fb80b30b

.github/ci_commit_pins/vision.txt

@@ -1 +1 @@
-d23a6e1664d20707c11781299611436e1f0c104f
+2c127da8b5e2e8f44b50994c6cb931bcca267cfe

.github/ci_commit_pins/xla.txt

@@ -1 +1 @@
-5ea4535f0699f366adb554183a65ebf7dc34a8be
+r2.3

.github/label_to_label.yml

@@ -1,13 +0,0 @@
-# Use this to auto apply labels based on other labels.  Applies to both PRs and
-# issues. Currently only supports any and all
-- any:
-  - "module: custom operators"
-  - "module: aotdispatch"
-  then:
-  - "module: pt2-dispatcher"
-- any:
-  - "module: dynamo"
-  - "module: pt2-dispatcher"
-  - "module: inductor"
-  then:
-  - "oncall: pt2"

.github/labeler.yml

@@ -35,9 +35,6 @@
 - test/distributed/tensor/parallel/test_fsdp_2d_parallel.py
 - torch/distributed/_tensor/**
 - torch/distributed/fsdp/**
-- torch/csrc/inductor/**
-- test/cpp/aoti_abi_check/**
-- test/cpp/aoti_inference/**
 
 "module: cpu":
 - aten/src/ATen/cpu/**
@@ -58,17 +55,6 @@
 - third_party/mkl-dnn.BUILD
 - torch/csrc/jit/codegen/onednn/**
 - test/test_jit_llga_fuser.py
-- test/test_mkldnn.py
-
-"ciflow/linux-aarch64":
-- third_party/ideep
-- caffe2/ideep/**
-- caffe2/python/ideep/**
-- cmake/Modules/FindMKLDNN.cmake
-- third_party/mkl-dnn.BUILD
-- torch/csrc/jit/codegen/onednn/**
-- test/test_jit_llga_fuser.py
-- test/test_mkldnn.py
 
 "module: amp (automated mixed precision)":
 - torch/amp/**

.github/lf-canary-scale-config.yml

@@ -1,281 +0,0 @@
-# Defines runner types that will be provisioned by by LF Self-hosted
-# runners for pytorch/pytorch-canary and their labels.
-#
-# Runners listed here will be available as self hosted runners.
-# Configuration is directly pulled from the main branch.
-#
-# Default values:
-#
-# runner_types:
-#   runner_label: # label to specify in the Github Actions workflow
-#     instance_type: m4.large
-#     os: linux
-#     max_available: 20
-#     disk_size: 50
-#     is_ephemeral: true
-
-runner_types:
-  lf.c.linux.12xlarge:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-  lf.c.linux.24xl.spr-metal:
-    disk_size: 200
-    instance_type: c7i.metal-24xl
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.c.linux.16xlarge.spr:
-    disk_size: 200
-    instance_type: c7i.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.c.linux.12xlarge.ephemeral:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: true
-    max_available: 300
-    os: linux
-  lf.c.linux.16xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.c.linux.24xlarge:
-    disk_size: 150
-    instance_type: c5.24xlarge
-    is_ephemeral: false
-    max_available: 250
-    os: linux
-  lf.c.linux.2xlarge:
-    disk_size: 150
-    instance_type: c5.2xlarge
-    is_ephemeral: false
-    max_available: 3120
-    os: linux
-  lf.c.linux.4xlarge:
-    disk_size: 150
-    instance_type: c5.4xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-  lf.c.linux.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.4xlarge
-    is_ephemeral: false
-    max_available: 520
-    os: linux
-  lf.c.linux.8xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.8xlarge
-    is_ephemeral: false
-    max_available: 400
-    os: linux
-  lf.c.linux.g4dn.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.12xlarge
-    is_ephemeral: false
-    max_available: 50
-    os: linux
-  lf.c.linux.g4dn.metal.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.metal
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.c.linux.g5.48xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.48xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-  lf.c.linux.g5.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.12xlarge
-    is_ephemeral: false
-    max_available: 150
-    os: linux
-  lf.c.linux.g5.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.4xlarge
-    is_ephemeral: false
-    max_available: 1200
-    os: linux
-  lf.c.linux.large:
-    disk_size: 15
-    instance_type: c5.large
-    is_ephemeral: false
-    os: linux
-  lf.c.linux.arm64.2xlarge:
-    disk_size: 256
-    instance_type: t4g.2xlarge
-    is_ephemeral: false
-    max_available: 200
-    os: linux
-  lf.c.linux.arm64.m7g.2xlarge:
-    disk_size: 256
-    instance_type: m7g.2xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-  lf.c.windows.4xlarge:
-    disk_size: 256
-    instance_type: c5d.4xlarge
-    is_ephemeral: true
-    max_available: 420
-    os: windows
-  lf.c.windows.4xlarge.nonephemeral:
-    disk_size: 256
-    instance_type: c5d.4xlarge
-    is_ephemeral: false
-    max_available: 420
-    os: windows
-  lf.c.windows.8xlarge.nvidia.gpu:
-    disk_size: 256
-    instance_type: p3.2xlarge
-    is_ephemeral: true
-    max_available: 150
-    os: windows
-  lf.c.windows.8xlarge.nvidia.gpu.nonephemeral:
-    disk_size: 256
-    instance_type: p3.2xlarge
-    is_ephemeral: false
-    max_available: 150
-    os: windows
-  lf.c.windows.g5.4xlarge.nvidia.gpu:
-    disk_size: 256
-    instance_type: g5.4xlarge
-    is_ephemeral: false
-    max_available: 250
-    os: windows
-
-  ### Setup runner types to test the Amazon Linux 2023 AMI
-  lf.c.amz2023.linux.12xlarge:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.24xl.spr-metal:
-    disk_size: 200
-    instance_type: c7i.metal-24xl
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.16xlarge.spr:
-    disk_size: 200
-    instance_type: c7i.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.12xlarge.ephemeral:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: true
-    max_available: 300
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.16xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.24xlarge:
-    disk_size: 150
-    instance_type: c5.24xlarge
-    is_ephemeral: false
-    max_available: 250
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.2xlarge:
-    disk_size: 150
-    instance_type: c5.2xlarge
-    is_ephemeral: false
-    max_available: 3120
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.4xlarge:
-    disk_size: 150
-    instance_type: c5.4xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.4xlarge
-    is_ephemeral: false
-    max_available: 520
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.8xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.8xlarge
-    is_ephemeral: false
-    max_available: 400
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.g4dn.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.12xlarge
-    is_ephemeral: false
-    max_available: 50
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.g4dn.metal.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.metal
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.g5.48xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.48xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.g5.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.12xlarge
-    is_ephemeral: false
-    max_available: 150
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.g5.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.4xlarge
-    is_ephemeral: false
-    max_available: 1200
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.large:
-    disk_size: 15
-    instance_type: c5.large
-    is_ephemeral: false
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.arm64.2xlarge:
-    disk_size: 256
-    instance_type: t4g.2xlarge
-    is_ephemeral: false
-    max_available: 200
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.c.amz2023.linux.arm64.m7g.2xlarge:
-    disk_size: 256
-    instance_type: m7g.2xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64

.github/lf-scale-config.yml

@@ -1,281 +0,0 @@
-# Defines runner types that will be provisioned by by LF Self-hosted
-# runners for pytorch/pytorch and their labels.
-#
-# Runners listed here will be available as self hosted runners.
-# Configuration is directly pulled from the main branch.
-#
-# Default values:
-#
-# runner_types:
-#   runner_label: # label to specify in the Github Actions workflow
-#     instance_type: m4.large
-#     os: linux
-#     max_available: 20
-#     disk_size: 50
-#     is_ephemeral: true
-
-runner_types:
-  lf.linux.12xlarge:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-  lf.linux.24xl.spr-metal:
-    disk_size: 200
-    instance_type: c7i.metal-24xl
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.linux.16xlarge.spr:
-    disk_size: 200
-    instance_type: c7i.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.linux.12xlarge.ephemeral:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: true
-    max_available: 300
-    os: linux
-  lf.linux.16xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.linux.24xlarge:
-    disk_size: 150
-    instance_type: c5.24xlarge
-    is_ephemeral: false
-    max_available: 250
-    os: linux
-  lf.linux.2xlarge:
-    disk_size: 150
-    instance_type: c5.2xlarge
-    is_ephemeral: false
-    max_available: 3120
-    os: linux
-  lf.linux.4xlarge:
-    disk_size: 150
-    instance_type: c5.4xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-  lf.linux.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.4xlarge
-    is_ephemeral: false
-    max_available: 520
-    os: linux
-  lf.linux.8xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.8xlarge
-    is_ephemeral: false
-    max_available: 400
-    os: linux
-  lf.linux.g4dn.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.12xlarge
-    is_ephemeral: false
-    max_available: 50
-    os: linux
-  lf.linux.g4dn.metal.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.metal
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-  lf.linux.g5.48xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.48xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-  lf.linux.g5.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.12xlarge
-    is_ephemeral: false
-    max_available: 150
-    os: linux
-  lf.linux.g5.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.4xlarge
-    is_ephemeral: false
-    max_available: 1200
-    os: linux
-  lf.linux.large:
-    disk_size: 15
-    instance_type: c5.large
-    is_ephemeral: false
-    os: linux
-  lf.linux.arm64.2xlarge:
-    disk_size: 256
-    instance_type: t4g.2xlarge
-    is_ephemeral: false
-    max_available: 200
-    os: linux
-  lf.linux.arm64.m7g.2xlarge:
-    disk_size: 256
-    instance_type: m7g.2xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-  lf.windows.4xlarge:
-    disk_size: 256
-    instance_type: c5d.4xlarge
-    is_ephemeral: true
-    max_available: 420
-    os: windows
-  lf.windows.4xlarge.nonephemeral:
-    disk_size: 256
-    instance_type: c5d.4xlarge
-    is_ephemeral: false
-    max_available: 420
-    os: windows
-  lf.windows.8xlarge.nvidia.gpu:
-    disk_size: 256
-    instance_type: p3.2xlarge
-    is_ephemeral: true
-    max_available: 150
-    os: windows
-  lf.windows.8xlarge.nvidia.gpu.nonephemeral:
-    disk_size: 256
-    instance_type: p3.2xlarge
-    is_ephemeral: false
-    max_available: 150
-    os: windows
-  lf.windows.g5.4xlarge.nvidia.gpu:
-    disk_size: 256
-    instance_type: g5.4xlarge
-    is_ephemeral: false
-    max_available: 250
-    os: windows
-
-  ### Setup runner types to test the Amazon Linux 2023 AMI
-  lf.amz2023.linux.12xlarge:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.24xl.spr-metal:
-    disk_size: 200
-    instance_type: c7i.metal-24xl
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.16xlarge.spr:
-    disk_size: 200
-    instance_type: c7i.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.12xlarge.ephemeral:
-    disk_size: 200
-    instance_type: c5.12xlarge
-    is_ephemeral: true
-    max_available: 300
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.16xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.16xlarge
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.24xlarge:
-    disk_size: 150
-    instance_type: c5.24xlarge
-    is_ephemeral: false
-    max_available: 250
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.2xlarge:
-    disk_size: 150
-    instance_type: c5.2xlarge
-    is_ephemeral: false
-    max_available: 3120
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.4xlarge:
-    disk_size: 150
-    instance_type: c5.4xlarge
-    is_ephemeral: false
-    max_available: 1000
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.4xlarge
-    is_ephemeral: false
-    max_available: 520
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.8xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g3.8xlarge
-    is_ephemeral: false
-    max_available: 400
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.g4dn.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.12xlarge
-    is_ephemeral: false
-    max_available: 50
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.g4dn.metal.nvidia.gpu:
-    disk_size: 150
-    instance_type: g4dn.metal
-    is_ephemeral: false
-    max_available: 30
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.g5.48xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.48xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.g5.12xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.12xlarge
-    is_ephemeral: false
-    max_available: 150
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.g5.4xlarge.nvidia.gpu:
-    disk_size: 150
-    instance_type: g5.4xlarge
-    is_ephemeral: false
-    max_available: 1200
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.large:
-    disk_size: 15
-    instance_type: c5.large
-    is_ephemeral: false
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.arm64.2xlarge:
-    disk_size: 256
-    instance_type: t4g.2xlarge
-    is_ephemeral: false
-    max_available: 200
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
-  lf.amz2023.linux.arm64.m7g.2xlarge:
-    disk_size: 256
-    instance_type: m7g.2xlarge
-    is_ephemeral: false
-    max_available: 20
-    os: linux
-    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64

.github/merge_rules.yaml

@@ -27,12 +27,13 @@
   - third_party/onnx
   - caffe2/python/onnx/**
   approved_by:
+  - BowenBao
+  - abock
   - justinchuby
-  - liqunfu
   - shubhambhokare1
+  - thiagocrepaldi
   - titaiwangms
   - wschin
-  - xadupre
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -235,25 +236,6 @@
   - Lint
   - pull
 
-- name: XPU ATen
-  patterns:
-  - aten/src/ATen/xpu/**
-  - c10/xpu/**
-  - torch/csrc/xpu/**
-  - torch/xpu/**
-  - test/xpu/**
-  - test/test_xpu.py
-  - third_party/xpu.txt
-  - .ci/docker/ci_commit_pins/triton-xpu.txt
-  approved_by:
-  - EikanWang
-  - jgong5
-  - gujinghui
-  mandatory_checks_name:
-  - EasyCLA
-  - Lint
-  - pull
-
 - name: Distributions
   patterns:
   - torch/distributions/**
@@ -286,7 +268,6 @@
   - test/cpp/dist_autograd/**
   - test/cpp/rpc/**
   approved_by:
-  - wconstab
   - mrshenli
   - pritamdamania87
   - zhaojuanmao
@@ -313,25 +294,6 @@
   - Lint
   - pull
 
-- name: DCP
-  patterns:
-  - torch/distributed/checkpoint/**
-  approved_by:
-  - LucasLLC
-  - fegin
-  - wz337
-  - saumishr
-  - daulet-askarov
-  - pradeepdfb
-  - kirtiteja
-  - mhorowitz
-  - saiteja64
-  mandatory_checks_name:
-  - EasyCLA
-  - Lint
-  - pull
-
-
 - name: IDEEP
   patterns:
   - third_party/ideep
@@ -395,22 +357,12 @@
 
 - name: CPU inductor
   patterns:
-  - torch/_inductor/mkldnn_ir.py
-  - torch/_inductor/mkldnn_lowerings.py
   - torch/_inductor/fx_passes/mkldnn_fusion.py
   - torch/_inductor/fx_passes/quantization.py
-  - torch/_inductor/codegen/cpp_prefix.h
   - torch/_inductor/codegen/cpp.py
-  - torch/_inductor/codegen/cpp_utils.py
-  - torch/_inductor/codegen/cpp_micro_gemm.py
-  - torch/_inductor/codegen/cpp_template_kernel.py
-  - torch/_inductor/codegen/cpp_template.py
-  - torch/_inductor/codegen/cpp_gemm_template.py
   - test/inductor/test_mkldnn_pattern_matcher.py
-  - test/inductor/test_cpu_repro.py
+  - test/inductor/test_cpu_repo.py
   - test/inductor/test_cpu_cpp_wrapper.py
-  - test/inductor/test_cpu_select_algorithm.py
-  - aten/src/ATen/cpu/**
   - aten/src/ATen/native/quantized/cpu/**
   - test/quantization/core/test_quantized_op.py
   - torch/ao/quantization/quantizer/x86_inductor_quantizer.py

.github/pytorch-probot.yml

@@ -7,9 +7,6 @@ ciflow_push_tags:
 - ciflow/binaries_wheel
 - ciflow/inductor
 - ciflow/inductor-perf-compare
-- ciflow/inductor-micro-benchmark
-- ciflow/inductor-cu124
-- ciflow/linux-aarch64
 - ciflow/mps
 - ciflow/nightly
 - ciflow/periodic
@@ -18,12 +15,9 @@ ciflow_push_tags:
 - ciflow/trunk
 - ciflow/unstable
 - ciflow/xpu
-- ciflow/torchbench
 retryable_workflows:
+- lint
 - pull
 - trunk
 - linux-binary
 - windows-binary
-labeler_config: labeler.yml
-label_to_label_config: label_to_label.yml
-mergebot: True

.github/requirements-gha-cache.txt

@@ -5,11 +5,11 @@
 #   functorch/docs/requirements.txt
 #   .ci/docker/requirements-ci.txt
 boto3==1.19.12
-jinja2==3.1.4
+jinja2==3.0.1
 lintrunner==0.10.7
 ninja==1.10.0.post1
 nvidia-ml-py==11.525.84
 pyyaml==6.0
-requests==2.32.2
+requests==2.31.0
 rich==10.9.0
 rockset==1.0.3

.github/requirements/conda-env-Linux-X64.txt

@@ -4,5 +4,6 @@ mkl-include=2022.1.0
 ninja=1.10.2
 numpy=1.23.3
 pyyaml=6.0
+requests=2.31.0
 setuptools=68.2.2
-typing-extensions=4.9.0
+typing-extensions=4.3.0

.github/requirements/conda-env-iOS.txt

@@ -3,5 +3,6 @@ cmake=3.22.1
 ninja=1.10.2
 numpy=1.23.3
 pyyaml=6.0
+requests=2.31.0
 setuptools=68.2.2
-typing-extensions=4.9.0
+typing-extensions=4.3.0

.github/requirements/conda-env-macOS-ARM64

@@ -2,7 +2,7 @@ numpy=1.22.3
 pyyaml=6.0
 setuptools=61.2.0
 cmake=3.22.*
-typing-extensions=4.9.0
+typing-extensions=4.3.0
 dataclasses=0.8
 pip=22.2.2
 pillow=10.0.1

.github/requirements/conda-env-macOS-X64

@@ -4,7 +4,7 @@ numpy=1.21.2
 pyyaml=5.3
 setuptools=46.0.0
 cmake=3.22.*
-typing-extensions=4.9.0
+typing-extensions=4.3.0
 dataclasses=0.8
 pip=22.2.2
 pillow=10.0.1

.github/requirements/pip-requirements-iOS.txt

@@ -1,4 +1,4 @@
 # iOS simulator requirements
 coremltools==5.0b5
 protobuf==3.20.2
-optree==0.12.1
+optree==0.9.1

.github/requirements/pip-requirements-macOS.txt

@@ -17,16 +17,13 @@ pytest-xdist==3.3.1
 pytest-rerunfailures==10.3
 pytest-flakefinder==1.1.0
 scipy==1.10.1
-sympy==1.12.1 ; python_version == "3.8"
-sympy>=1.13.0 ; python_version >= "3.9"
+sympy==1.11.1
 unittest-xml-reporting<=3.2.0,>=2.0.0
 xdoctest==1.1.0
 filelock==3.6.0
+sympy==1.11.1
 pytest-cpp==2.3.0
 rockset==1.0.3
 z3-solver==4.12.2.0
 tensorboard==2.13.0
-optree==0.12.1
-# NB: test_hparams_* from test_tensorboard is failing with protobuf 5.26.0 in
-# which the stringify metadata is wrong when escaping double quote
-protobuf==3.20.2
+optree==0.9.1

.github/scripts/amd/package_triton_wheel.sh

@@ -1,101 +0,0 @@
-set -ex
-
-# Set ROCM_HOME isn't available, use ROCM_PATH if set or /opt/rocm
-ROCM_HOME="${ROCM_HOME:-${ROCM_PATH:-/opt/rocm}}"
-
-# Find rocm_version.h header file for ROCm version extract
-rocm_version_h="${ROCM_HOME}/include/rocm-core/rocm_version.h"
-if [ ! -f "$rocm_version_h" ]; then
-    rocm_version_h="${ROCM_HOME}/include/rocm_version.h"
-fi
-
-# Error out if rocm_version.h not found
-if [ ! -f "$rocm_version_h" ]; then
-    echo "Error: rocm_version.h not found in expected locations." >&2
-    exit 1
-fi
-
-# Extract major, minor and patch ROCm version numbers
-MAJOR_VERSION=$(grep 'ROCM_VERSION_MAJOR' "$rocm_version_h" | awk '{print $3}')
-MINOR_VERSION=$(grep 'ROCM_VERSION_MINOR' "$rocm_version_h" | awk '{print $3}')
-PATCH_VERSION=$(grep 'ROCM_VERSION_PATCH' "$rocm_version_h" | awk '{print $3}')
-ROCM_INT=$(($MAJOR_VERSION * 10000 + $MINOR_VERSION * 100 + $PATCH_VERSION))
-echo "ROCm version: $ROCM_INT"
-
-# Check TRITON_ROCM_DIR is set
-if [[ -z "${TRITON_ROCM_DIR}" ]]; then
-    export TRITON_ROCM_DIR=third_party/amd/backend
-fi
-
-# Remove packaged libs and headers
-rm -rf $TRITON_ROCM_DIR/include/*
-
-LIBTINFO_PATH="/usr/lib64/libtinfo.so.5"
-LIBNUMA_PATH="/usr/lib64/libnuma.so.1"
-LIBELF_PATH="/usr/lib64/libelf.so.1"
-
-OS_SO_PATHS=(
-    $LIBELF_PATH
-    $LIBNUMA_PATH
-    $LIBTINFO_PATH
-)
-
-for lib in "${OS_SO_PATHS[@]}"
-do
-    cp $lib $TRITON_ROCM_DIR/lib/
-done
-
-# Required ROCm libraries
-if [[ "${MAJOR_VERSION}" == "6" ]]; then
-    libamdhip="libamdhip64.so.6"
-else
-    libamdhip="libamdhip64.so.5"
-fi
-
-# Required ROCm libraries - ROCm 6.0
-ROCM_SO=(
-    "${libamdhip}"
-    "libhsa-runtime64.so.1"
-    "libamd_comgr.so.2"
-    "libdrm.so.2"
-    "libdrm_amdgpu.so.1"
-)
-
-if [[ $ROCM_INT -ge 60100 ]]; then
-    ROCM_SO+=("librocprofiler-register.so.0")
-fi
-
-for lib in "${ROCM_SO[@]}"
-do
-    file_path=($(find $ROCM_HOME/lib/ -name "$lib")) # First search in lib
-    if [[ -z $file_path ]]; then
-        if [ -d "$ROCM_HOME/lib64/" ]; then
-            file_path=($(find $ROCM_HOME/lib64/ -name "$lib")) # Then search in lib64
-        fi
-    fi
-    if [[ -z $file_path ]]; then
-        file_path=($(find $ROCM_HOME/ -name "$lib")) # Then search in ROCM_HOME
-    fi
-    if [[ -z $file_path ]]; then
-        file_path=($(find /opt/ -name "$lib")) # Then search in /opt
-    fi
-    if [[ -z $file_path ]]; then
-            echo "Error: Library file $lib is not found." >&2
-            exit 1
-    fi
-
-    cp $file_path $TRITON_ROCM_DIR/lib
-    # When running locally, and not building a wheel, we need to satisfy shared objects requests that don't look for versions
-    LINKNAME=$(echo $lib | sed -e 's/\.so.*/.so/g')
-    ln -sf $lib $TRITON_ROCM_DIR/lib/$LINKNAME
-
-done
-
-# Copy Include Files
-cp -r $ROCM_HOME/include/hip $TRITON_ROCM_DIR/include
-cp -r $ROCM_HOME/include/roctracer $TRITON_ROCM_DIR/include
-cp -r $ROCM_HOME/include/hsa $TRITON_ROCM_DIR/include
-
-# Copy linker
-mkdir -p $TRITON_ROCM_DIR/llvm/bin
-cp $ROCM_HOME/llvm/bin/ld.lld $TRITON_ROCM_DIR/llvm/bin/

.github/scripts/amd/patch_triton_wheel.sh

@@ -1,103 +0,0 @@
-#!/bin/bash
-set -x
-
-if [ -z "$1" ]; then
-    echo "Need wheel location argument" && exit 1
-fi
-
-WHEELHOUSE_DIR=$1
-PATCHELF_BIN=patchelf
-ROCM_LIB=backends/amd/lib
-ROCM_LD=backends/amd/llvm/bin
-PREFIX=triton
-fname_without_so_number() {
-    LINKNAME=$(echo $1 | sed -e 's/\.so.*/.so/g')
-    echo "$LINKNAME"
-}
-
-replace_needed_sofiles() {
-    find $1 -name '*.so*' -o -name 'ld.lld' | while read sofile; do
-        origname=$2
-        patchedname=$3
-        if [[ "$origname" != "$patchedname" ]]; then
-            set +e
-            origname=$($PATCHELF_BIN --print-needed $sofile | grep "$origname.*")
-            ERRCODE=$?
-            set -e
-            if [ "$ERRCODE" -eq "0" ]; then
-                echo "patching $sofile entry $origname to $patchedname"
-                $PATCHELF_BIN --replace-needed $origname $patchedname $sofile
-            fi
-        fi
-    done
-}
-
-mkdir  -p "/tmp_dir"
-pushd /tmp_dir
-for pkg in /$WHEELHOUSE_DIR/*triton*.whl; do
-    echo "Modifying $pkg"
-    rm -rf tmp
-    mkdir -p tmp
-    cd tmp
-    cp $pkg .
-    unzip -q $(basename $pkg)
-    rm -f $(basename $pkg)
-    $PATCHELF_BIN --set-rpath ${LD_SO_RPATH:-'$ORIGIN:$ORIGIN/../../lib'} $PREFIX/$ROCM_LD/ld.lld
-    $PATCHELF_BIN --print-rpath $PREFIX/$ROCM_LD/ld.lld
-    # Modify libtriton.so as it sits in _C directory apart from its dependencies
-    find $PREFIX/_C -type f -name "*.so*" | while read sofile; do
-        echo "Setting rpath of $sofile"
-        $PATCHELF_BIN --set-rpath ${C_SO_RPATH:-'$ORIGIN:$ORIGIN/'../$ROCM_LIB} ${FORCE_RPATH:-} $sofile
-        $PATCHELF_BIN --print-rpath $sofile
-    done
-
-    # All included dependencies are included in a single lib directory
-    deps=()
-    deps_soname=()
-    while read sofile; do
-        echo "Setting rpath of $sofile to ${LIB_SO_RPATH:-'$ORIGIN'}"
-        $PATCHELF_BIN --set-rpath ${LIB_SO_RPATH:-'$ORIGIN'} ${FORCE_RPATH:-} $sofile
-        $PATCHELF_BIN --print-rpath $sofile
-        deps+=("$sofile")
-        deps_soname+=("$(basename $sofile)")
-    done < <(find $PREFIX/$ROCM_LIB -type f -name "*.so*")
-
-    patched=()
-    for filepath in "${deps[@]}"; do
-        filename=$(basename $filepath)
-        destpath=$PREFIX/$ROCM_LIB/$filename
-        if [[ "$filepath" != "$destpath" ]]; then
-            cp $filepath $destpath
-        fi
-        patchedpath=$(fname_without_so_number $destpath)
-        patchedname=$(basename $patchedpath)
-        if [[ "$destpath" != "$patchedpath" ]]; then
-            mv $destpath $patchedpath
-        fi
-        patched+=("$patchedname")
-        echo "Copied $filepath to $patchedpath"
-    done
-
-    # Go through all required shared objects and see if any of our other objects are dependants.  If so, replace so.ver wth so
-    for ((i=0;i<${#deps[@]};++i)); do
-        echo "replacing "${deps_soname[i]} ${patched[i]}
-        replace_needed_sofiles $PREFIX/$ROCM_LIB ${deps_soname[i]} ${patched[i]}
-        replace_needed_sofiles $PREFIX/_C ${deps_soname[i]} ${patched[i]}
-        replace_needed_sofiles $PREFIX/$ROCM_LD ${deps_soname[i]} ${patched[i]}
-    done
-
-    # Re-bundle whl with so adjustments
-    zip -rqy $(basename $pkg) *
-
-    if [[ -z "${MANYLINUX_VERSION}" ]]; then
-        newpkg=$pkg
-    else
-        newpkg=$(echo $pkg | sed -e "s/\linux_x86_64/${MANYLINUX_VERSION}/g")
-    fi
-
-    # Remove original whl
-    rm -f $pkg
-
-    # Move rebuilt whl to original location with new name.
-    mv $(basename $pkg) $newpkg
-done

.github/scripts/build_triton_wheel.py

@@ -1,5 +1,4 @@
 #!/usr/bin/env python3
-
 import os
 import shutil
 import sys
@@ -8,17 +7,15 @@ from subprocess import check_call
 from tempfile import TemporaryDirectory
 from typing import Optional
 
-
 SCRIPT_DIR = Path(__file__).parent
 REPO_DIR = SCRIPT_DIR.parent.parent
 
+# TODO: Remove me once Triton version is again in sync for vanilla and ROCm
+ROCM_TRITION_VERSION = "2.2.0"
 
-def read_triton_pin(device: str = "cuda") -> str:
-    triton_file = "triton.txt"
-    if device == "rocm":
-        triton_file = "triton-rocm.txt"
-    elif device == "xpu":
-        triton_file = "triton-xpu.txt"
+
+def read_triton_pin(rocm_hash: bool = False) -> str:
+    triton_file = "triton.txt" if not rocm_hash else "triton-rocm.txt"
     with open(REPO_DIR / ".ci" / "docker" / "ci_commit_pins" / triton_file) as f:
         return f.read().strip()
 
@@ -35,6 +32,27 @@ def check_and_replace(inp: str, src: str, dst: str) -> str:
     return inp.replace(src, dst)
 
 
+def patch_setup_py(
+    path: Path,
+    *,
+    version: str,
+    name: str = "triton",
+    expected_version: Optional[str] = None,
+) -> None:
+    with open(path) as f:
+        orig = f.read()
+    # Replace name
+    orig = check_and_replace(orig, 'name="triton",', f'name="{name}",')
+    # Replace version
+    if not expected_version:
+        expected_version = read_triton_version()
+    orig = check_and_replace(
+        orig, f'version="{expected_version}",', f'version="{version}",'
+    )
+    with open(path, "w") as f:
+        f.write(orig)
+
+
 def patch_init_py(
     path: Path, *, version: str, expected_version: Optional[str] = None
 ) -> None:
@@ -55,7 +73,7 @@ def build_triton(
     version: str,
     commit_hash: str,
     build_conda: bool = False,
-    device: str = "cuda",
+    build_rocm: bool = False,
     py_version: Optional[str] = None,
     release: bool = False,
 ) -> Path:
@@ -74,20 +92,23 @@ def build_triton(
     with TemporaryDirectory() as tmpdir:
         triton_basedir = Path(tmpdir) / "triton"
         triton_pythondir = triton_basedir / "python"
-        triton_repo = "https://github.com/openai/triton"
-        if device == "rocm":
+        if build_rocm:
+            triton_repo = "https://github.com/ROCmSoftwarePlatform/triton"
             triton_pkg_name = "pytorch-triton-rocm"
-        elif device == "xpu":
-            triton_pkg_name = "pytorch-triton-xpu"
-            triton_repo = "https://github.com/intel/intel-xpu-backend-for-triton"
         else:
+            triton_repo = "https://github.com/openai/triton"
             triton_pkg_name = "pytorch-triton"
-        check_call(["git", "clone", triton_repo, "triton"], cwd=tmpdir)
+        check_call(["git", "clone", triton_repo], cwd=tmpdir)
         if release:
             ver, rev, patch = version.split(".")
-            check_call(
-                ["git", "checkout", f"release/{ver}.{rev}.x"], cwd=triton_basedir
-            )
+            if build_rocm:
+                check_call(
+                    ["git", "checkout", f"release/2.2.x"], cwd=triton_basedir
+                )
+            else:
+                check_call(
+                    ["git", "checkout", f"release/{ver}.{rev}.x"], cwd=triton_basedir
+                )
         else:
             check_call(["git", "checkout", commit_hash], cwd=triton_basedir)
 
@@ -100,7 +121,7 @@ def build_triton(
                 print("source:\n  path: .\n", file=meta)
                 print(
                     "build:\n  string: py{{py}}\n  number: 1\n  script: cd python; "
-                    "python setup.py install --record=record.txt\n",
+                    "python setup.py install --single-version-externally-managed --record=record.txt\n",
                     " script_env:\n   - MAX_JOBS\n",
                     file=meta,
                 )
@@ -146,15 +167,18 @@ def build_triton(
         patch_init_py(
             triton_pythondir / "triton" / "__init__.py",
             version=f"{version}",
-            expected_version=None,
+            expected_version=ROCM_TRITION_VERSION if build_rocm else None,
         )
 
-        if device == "rocm":
-            check_call(
-                [f"{SCRIPT_DIR}/amd/package_triton_wheel.sh"],
-                cwd=triton_basedir,
-                shell=True,
+        if build_rocm:
+            # TODO: Remove me when ROCM triton is updated
+            patch_setup_py(
+                triton_pythondir / "setup.py",
+                name=triton_pkg_name,
+                version=f"{version}",
+                expected_version=ROCM_TRITION_VERSION,
             )
+            check_call("scripts/amd/setup_rocm_libs.sh", cwd=triton_basedir, shell=True)
             print("ROCm libraries setup for triton installation...")
 
         check_call(
@@ -164,11 +188,8 @@ def build_triton(
         whl_path = next(iter((triton_pythondir / "dist").glob("*.whl")))
         shutil.copy(whl_path, Path.cwd())
 
-        if device == "rocm":
-            check_call(
-                [f"{SCRIPT_DIR}/amd/patch_triton_wheel.sh", Path.cwd()],
-                cwd=triton_basedir,
-            )
+        if build_rocm:
+            check_call("scripts/amd/fix_so.sh", cwd=triton_basedir, shell=True)
 
         return Path.cwd() / whl_path.name
 
@@ -179,19 +200,17 @@ def main() -> None:
     parser = ArgumentParser("Build Triton binaries")
     parser.add_argument("--release", action="store_true")
     parser.add_argument("--build-conda", action="store_true")
-    parser.add_argument(
-        "--device", type=str, default="cuda", choices=["cuda", "rocm", "xpu"]
-    )
+    parser.add_argument("--build-rocm", action="store_true")
     parser.add_argument("--py-version", type=str)
     parser.add_argument("--commit-hash", type=str)
     parser.add_argument("--triton-version", type=str, default=read_triton_version())
     args = parser.parse_args()
 
     build_triton(
-        device=args.device,
+        build_rocm=args.build_rocm,
         commit_hash=args.commit_hash
         if args.commit_hash
-        else read_triton_pin(args.device),
+        else read_triton_pin(args.build_rocm),
         version=args.triton_version,
         build_conda=args.build_conda,
         py_version=args.py_version,

.github/scripts/check_labels.py

@@ -5,6 +5,7 @@ import sys
 from typing import Any
 
 from github_utils import gh_delete_comment, gh_post_pr_comment
+
 from gitutils import get_git_remote_name, get_git_repo_dir, GitRepo
 from label_utils import has_required_labels, is_label_err_comment, LABEL_ERR_MSG
 from trymerge import GitHubPR

.github/scripts/cherry_pick.py

@@ -3,10 +3,12 @@
 import json
 import os
 import re
-from typing import Any, cast, Dict, List, Optional
+from typing import Any, Optional
+
 from urllib.error import HTTPError
 
-from github_utils import gh_fetch_url, gh_post_pr_comment, gh_query_issues_by_labels
+from github_utils import gh_fetch_url, gh_post_pr_comment
+
 from gitutils import get_git_remote_name, get_git_repo_dir, GitRepo
 from trymerge import get_pr_commit_sha, GitHubPR
 
@@ -17,7 +19,6 @@ REQUIRES_ISSUE = {
     "critical",
     "fixnewfeature",
 }
-RELEASE_BRANCH_REGEX = re.compile(r"release/(?P<version>.+)")
 
 
 def parse_args() -> Any:
@@ -28,7 +29,7 @@ def parse_args() -> Any:
         "--onto-branch", type=str, required=True, help="the target release branch"
     )
     parser.add_argument(
-        "--github-actor", type=str, required=True, help="all the world's a stage"
+        "--github-actor", type=str, required=True, help="all the world’s a stage"
     )
     parser.add_argument(
         "--classification",
@@ -57,33 +58,6 @@ def get_merge_commit_sha(repo: GitRepo, pr: GitHubPR) -> Optional[str]:
     return commit_sha if pr.is_closed() else None
 
 
-def get_release_version(onto_branch: str) -> Optional[str]:
-    """
-    Return the release version if the target branch is a release branch
-    """
-    m = re.match(RELEASE_BRANCH_REGEX, onto_branch)
-    return m.group("version") if m else ""
-
-
-def get_tracker_issues(
-    org: str, project: str, onto_branch: str
-) -> List[Dict[str, Any]]:
-    """
-    Find the tracker issue from the repo. The tracker issue needs to have the title
-    like [VERSION] Release Tracker following the convention on PyTorch
-    """
-    version = get_release_version(onto_branch)
-    if not version:
-        return []
-
-    tracker_issues = gh_query_issues_by_labels(org, project, labels=["release tracker"])
-    if not tracker_issues:
-        return []
-
-    # Figure out the tracker issue from the list by looking at the title
-    return [issue for issue in tracker_issues if version in issue.get("title", "")]
-
-
 def cherry_pick(
     github_actor: str,
     repo: GitRepo,
@@ -103,49 +77,17 @@ def cherry_pick(
     )
 
     try:
-        org, project = repo.gh_owner_and_name()
-
-        cherry_pick_pr = ""
         if not dry_run:
+            org, project = repo.gh_owner_and_name()
             cherry_pick_pr = submit_pr(repo, pr, cherry_pick_branch, onto_branch)
 
-        tracker_issues_comments = []
-        tracker_issues = get_tracker_issues(org, project, onto_branch)
-        for issue in tracker_issues:
-            issue_number = int(str(issue.get("number", "0")))
-            if not issue_number:
-                continue
+            msg = f"The cherry pick PR is at {cherry_pick_pr}"
+            if fixes:
+                msg += f" and it is linked with issue {fixes}"
+            elif classification in REQUIRES_ISSUE:
+                msg += f" and it is recommended to link a {classification} cherry pick PR with an issue"
 
-            res = cast(
-                Dict[str, Any],
-                post_tracker_issue_comment(
-                    org,
-                    project,
-                    issue_number,
-                    pr.pr_num,
-                    cherry_pick_pr,
-                    classification,
-                    fixes,
-                    dry_run,
-                ),
-            )
-
-            comment_url = res.get("html_url", "")
-            if comment_url:
-                tracker_issues_comments.append(comment_url)
-
-        msg = f"The cherry pick PR is at {cherry_pick_pr}"
-        if fixes:
-            msg += f" and it is linked with issue {fixes}."
-        elif classification in REQUIRES_ISSUE:
-            msg += f" and it is recommended to link a {classification} cherry pick PR with an issue."
-
-        if tracker_issues_comments:
-            msg += " The following tracker issues are updated:\n"
-            for tracker_issues_comment in tracker_issues_comments:
-                msg += f"* {tracker_issues_comment}\n"
-
-        post_pr_comment(org, project, pr.pr_num, msg, dry_run)
+            post_comment(org, project, pr.pr_num, msg)
 
     finally:
         if current_branch:
@@ -217,9 +159,7 @@ def submit_pr(
         raise RuntimeError(msg) from error
 
 
-def post_pr_comment(
-    org: str, project: str, pr_num: int, msg: str, dry_run: bool = False
-) -> List[Dict[str, Any]]:
+def post_comment(org: str, project: str, pr_num: int, msg: str) -> None:
     """
     Post a comment on the PR itself to point to the cherry picking PR when success
     or print the error when failure
@@ -242,35 +182,7 @@ def post_pr_comment(
     comment = "\n".join(
         (f"### Cherry picking #{pr_num}", f"{msg}", "", f"{internal_debugging}")
     )
-    return gh_post_pr_comment(org, project, pr_num, comment, dry_run)
-
-
-def post_tracker_issue_comment(
-    org: str,
-    project: str,
-    issue_num: int,
-    pr_num: int,
-    cherry_pick_pr: str,
-    classification: str,
-    fixes: str,
-    dry_run: bool = False,
-) -> List[Dict[str, Any]]:
-    """
-    Post a comment on the tracker issue (if any) to record the cherry pick
-    """
-    comment = "\n".join(
-        (
-            "Link to landed trunk PR (if applicable):",
-            f"* https://github.com/{org}/{project}/pull/{pr_num}",
-            "",
-            "Link to release branch PR:",
-            f"* {cherry_pick_pr}",
-            "",
-            "Criteria Category:",
-            " - ".join((classification.capitalize(), fixes.capitalize())),
-        )
-    )
-    return gh_post_pr_comment(org, project, issue_num, comment, dry_run)
+    gh_post_pr_comment(org, project, pr_num, comment)
 
 
 def main() -> None:
@@ -302,7 +214,7 @@ def main() -> None:
 
     except RuntimeError as error:
         if not args.dry_run:
-            post_pr_comment(org, project, pr_num, str(error))
+            post_comment(org, project, pr_num, str(error))
         else:
             raise error
 

.github/scripts/close_nonexistent_disable_issues.py

@@ -10,7 +10,6 @@ import requests
 import rockset  # type: ignore[import]
 from gitutils import retries_decorator
 
-
 LOGS_QUERY = """
 with
     shas as (

.github/scripts/collect_ciflow_labels.py

@@ -1,12 +1,10 @@
 #!/usr/bin/env python3
-
 import sys
 from pathlib import Path
 from typing import Any, cast, Dict, List, Set
 
 import yaml
 
-
 GITHUB_DIR = Path(__file__).parent.parent
 
 

.github/scripts/comment_on_pr.py

@@ -23,10 +23,8 @@ def main() -> None:
 
     job_link = f"[job]({run_url})" if run_url is not None else "job"
     msg = (
-        f"The {args.action} {job_link} was canceled or timed out. This most often happen if two merge requests were issued"
-        + " for the same PR, or if merge job was waiting for more than 6 hours for tests to finish."
-        + " In later case, please do not hesitate to reissue the merge command\n"
-        + f" For more information see [pytorch-bot wiki]({BOT_COMMANDS_WIKI})."
+        f"The {args.action} {job_link} was canceled. If you believe this is a mistake,"
+        + f" then you can re trigger it through [pytorch-bot]({BOT_COMMANDS_WIKI})."
     )
 
     gh_post_pr_comment(org, project, args.pr_num, msg)

.github/scripts/convert_lintrunner_annotations_to_github.py

@@ -1,6 +1,7 @@
 import json
 import subprocess
 import sys
+
 from enum import Enum
 from pathlib import Path
 from typing import NamedTuple, Optional

.github/scripts/delete_old_branches.py

@@ -2,14 +2,12 @@
 import os
 import re
 from datetime import datetime
-from functools import lru_cache
 from pathlib import Path
 from typing import Any, Callable, Dict, List, Set
 
 from github_utils import gh_fetch_json_dict, gh_graphql
 from gitutils import GitRepo
 
-
 SEC_IN_DAY = 24 * 60 * 60
 CLOSED_PR_RETENTION = 30 * SEC_IN_DAY
 NO_PR_RETENTION = 1.5 * 365 * SEC_IN_DAY
@@ -20,7 +18,7 @@ ESTIMATED_TOKENS = [0]
 
 TOKEN = os.environ["GITHUB_TOKEN"]
 if not TOKEN:
-    raise Exception("GITHUB_TOKEN is not set")  # noqa: TRY002
+    raise Exception("GITHUB_TOKEN is not set")
 
 REPO_ROOT = Path(__file__).parent.parent.parent
 
@@ -189,17 +187,6 @@ def get_recent_prs() -> Dict[str, Any]:
     return prs_by_branch_base
 
 
-@lru_cache(maxsize=1)
-def get_open_prs() -> List[Dict[str, Any]]:
-    return paginate_graphql(
-        GRAPHQL_OPEN_PRS,
-        {"owner": "pytorch", "repo": "pytorch"},
-        lambda data: False,
-        lambda res: res["data"]["repository"]["pullRequests"]["nodes"],
-        lambda res: res["data"]["repository"]["pullRequests"]["pageInfo"],
-    )
-
-
 def get_branches_with_magic_label_or_open_pr() -> Set[str]:
     pr_infos: List[Dict[str, Any]] = paginate_graphql(
         GRAPHQL_NO_DELETE_BRANCH_LABEL,
@@ -209,7 +196,15 @@ def get_branches_with_magic_label_or_open_pr() -> Set[str]:
         lambda res: res["data"]["repository"]["label"]["pullRequests"]["pageInfo"],
     )
 
-    pr_infos.extend(get_open_prs())
+    pr_infos.extend(
+        paginate_graphql(
+            GRAPHQL_OPEN_PRS,
+            {"owner": "pytorch", "repo": "pytorch"},
+            lambda data: False,
+            lambda res: res["data"]["repository"]["pullRequests"]["nodes"],
+            lambda res: res["data"]["repository"]["pullRequests"]["pageInfo"],
+        )
+    )
 
     # Get the most recent PR for each branch base (group gh together)
     branch_bases = set()
@@ -275,41 +270,5 @@ def delete_branches() -> None:
         delete_branch(git_repo, branch)
 
 
-def delete_old_ciflow_tags() -> None:
-    # Deletes ciflow tags if they are associated with a closed PR or a specific
-    # commit.  Lightweight tags don't have information about the date they were
-    # created, so we can't check how old they are.  The script just assumes that
-    # ciflow tags should be deleted regardless of creation date.
-    git_repo = GitRepo(str(REPO_ROOT), "origin", debug=True)
-
-    def delete_tag(tag: str) -> None:
-        print(f"Deleting tag {tag}")
-        ESTIMATED_TOKENS[0] += 1
-        delete_branch(git_repo, f"refs/tags/{tag}")
-
-    tags = git_repo._run_git("tag").splitlines()
-    open_pr_numbers = [x["number"] for x in get_open_prs()]
-
-    for tag in tags:
-        try:
-            if ESTIMATED_TOKENS[0] > 400:
-                print("Estimated tokens exceeded, exiting")
-                break
-            if not tag.startswith("ciflow/"):
-                continue
-            re_match_pr = re.match(r"^ciflow\/.*\/(\d{5,6})$", tag)
-            re_match_sha = re.match(r"^ciflow\/.*\/([0-9a-f]{40})$", tag)
-            if re_match_pr:
-                pr_number = int(re_match_pr.group(1))
-                if pr_number in open_pr_numbers:
-                    continue
-                delete_tag(tag)
-            elif re_match_sha:
-                delete_tag(tag)
-        except Exception as e:
-            print(f"Failed to check tag {tag}: {e}")
-
-
 if __name__ == "__main__":
     delete_branches()
-    delete_old_ciflow_tags()

.github/scripts/docathon-label-sync.py

@@ -1,52 +0,0 @@
-import os
-import re
-import sys
-
-from github import Github
-
-
-def main() -> None:
-    token = os.environ.get("GITHUB_TOKEN")
-
-    repo_owner = "pytorch"
-    repo_name = "pytorch"
-    pull_request_number = int(sys.argv[1])
-
-    g = Github(token)
-    repo = g.get_repo(f"{repo_owner}/{repo_name}")
-    pull_request = repo.get_pull(pull_request_number)
-    pull_request_body = pull_request.body
-    # PR without description
-    if pull_request_body is None:
-        return
-
-    # get issue number from the PR body
-    if not re.search(r"#\d{1,6}", pull_request_body):
-        print("The pull request does not mention an issue.")
-        return
-    issue_number = int(re.findall(r"#(\d{1,6})", pull_request_body)[0])
-    issue = repo.get_issue(issue_number)
-    issue_labels = issue.labels
-    docathon_label_present = any(
-        label.name == "docathon-h1-2024" for label in issue_labels
-    )
-
-    # if the issue has a docathon label, add all labels from the issue to the PR.
-    if not docathon_label_present:
-        print("The 'docathon-h1-2024' label is not present in the issue.")
-        return
-    pull_request_labels = pull_request.get_labels()
-    pull_request_label_names = [label.name for label in pull_request_labels]
-    issue_label_names = [label.name for label in issue_labels]
-    labels_to_add = [
-        label for label in issue_label_names if label not in pull_request_label_names
-    ]
-    if not labels_to_add:
-        print("The pull request already has the same labels.")
-        return
-    pull_request.add_to_labels(*labels_to_add)
-    print("Labels added to the pull request!")
-
-
-if __name__ == "__main__":
-    main()

.github/scripts/ensure_actions_will_cancel.py

@@ -1,6 +1,7 @@
 #!/usr/bin/env python3
 
 import sys
+
 from pathlib import Path
 
 import yaml

.github/scripts/export_pytorch_labels.py

@@ -14,6 +14,7 @@ import json
 from typing import Any
 
 import boto3  # type: ignore[import]
+
 from label_utils import gh_get_labels
 
 

.github/scripts/filter_test_configs.py

@@ -1,7 +1,6 @@
 #!/usr/bin/env python3
 
 import json
-import logging
 import os
 import re
 import subprocess
@@ -9,18 +8,42 @@ import sys
 import warnings
 from enum import Enum
 from functools import lru_cache
-from logging import info
 from typing import Any, Callable, Dict, List, Optional, Set
 from urllib.request import Request, urlopen
 
 import yaml
 
-
 REENABLE_TEST_REGEX = "(?i)(Close(d|s)?|Resolve(d|s)?|Fix(ed|es)?) (#|https://github.com/pytorch/pytorch/issues/)([0-9]+)"
 
 PREFIX = "test-config/"
 
-logging.basicConfig(level=logging.INFO)
+# Same as shard names
+VALID_TEST_CONFIG_LABELS = {
+    f"{PREFIX}{label}"
+    for label in {
+        "backwards_compat",
+        "crossref",
+        "default",
+        "deploy",
+        "distributed",
+        "docs_tests",
+        "dynamo",
+        "force_on_cpu",
+        "functorch",
+        "inductor",
+        "inductor_distributed",
+        "inductor_huggingface",
+        "inductor_timm",
+        "inductor_torchbench",
+        "jit_legacy",
+        "multigpu",
+        "nogpu_AVX512",
+        "nogpu_NO_AVX2",
+        "slow",
+        "tsan",
+        "xla",
+    }
+}
 
 
 def is_cuda_or_rocm_job(job_name: Optional[str]) -> bool:
@@ -39,9 +62,9 @@ SUPPORTED_PERIODICAL_MODES: Dict[str, Callable[[Optional[str]], bool]] = {
 }
 
 # The link to the published list of disabled jobs
-DISABLED_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/disabled-jobs.json"
+DISABLED_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/disabled-jobs.json?versionId=qO7aEr.Og33PtLXfNq0j0yj.bbLC7SzR"
 # and unstable jobs
-UNSTABLE_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/unstable-jobs.json"
+UNSTABLE_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/unstable-jobs.json?versionId=7NhgpqKTtGXVUnL1C79KboTW_5qQx8y5"
 
 # Some constants used to handle disabled and unstable jobs
 JOB_NAME_SEP = "/"
@@ -67,12 +90,6 @@ def parse_args() -> Any:
     parser.add_argument(
         "--test-matrix", type=str, required=True, help="the original test matrix"
     )
-    parser.add_argument(
-        "--selected-test-configs",
-        type=str,
-        default="",
-        help="a comma-separated list of test configurations from the test matrix to keep",
-    )
     parser.add_argument(
         "--workflow", type=str, help="the name of the current workflow, i.e. pull"
     )
@@ -138,25 +155,19 @@ def get_labels(pr_number: int) -> Set[str]:
     }
 
 
-def filter_labels(labels: Set[str], label_regex: Any) -> Set[str]:
-    """
-    Return the list of matching labels
-    """
-    return {l for l in labels if re.match(label_regex, l)}
-
-
 def filter(test_matrix: Dict[str, List[Any]], labels: Set[str]) -> Dict[str, List[Any]]:
     """
     Select the list of test config to run from the test matrix. The logic works
     as follows:
 
-    If the PR has one or more test-config labels as specified, only these test configs
-    will be selected.  This also works with ciflow labels, for example, if a PR has both
-    ciflow/trunk and test-config/functorch, only trunk functorch builds and tests will
-    be run.
+    If the PR has one or more labels as specified in the VALID_TEST_CONFIG_LABELS set, only
+    these test configs will be selected.  This also works with ciflow labels, for example,
+    if a PR has both ciflow/trunk and test-config/functorch, only trunk functorch builds
+    and tests will be run
 
     If the PR has none of the test-config label, all tests are run as usual.
     """
+
     filtered_test_matrix: Dict[str, List[Any]] = {"include": []}
 
     for entry in test_matrix.get("include", []):
@@ -166,46 +177,23 @@ def filter(test_matrix: Dict[str, List[Any]], labels: Set[str]) -> Dict[str, Lis
 
         label = f"{PREFIX}{config_name.strip()}"
         if label in labels:
-            msg = f"Select {config_name} because label {label} is present in the pull request by the time the test starts"
-            info(msg)
+            print(
+                f"Select {config_name} because label {label} is presented in the pull request by the time the test starts"
+            )
             filtered_test_matrix["include"].append(entry)
 
-    test_config_labels = filter_labels(labels, re.compile(f"{PREFIX}.+"))
-    if not filtered_test_matrix["include"] and not test_config_labels:
-        info("Found no test-config label on the PR, so all test configs are included")
-        # Found no test-config label and the filtered test matrix is empty, return the same
+    valid_test_config_labels = labels.intersection(VALID_TEST_CONFIG_LABELS)
+
+    if not filtered_test_matrix["include"] and not valid_test_config_labels:
+        # Found no valid label and the filtered test matrix is empty, return the same
         # test matrix as before so that all tests can be run normally
         return test_matrix
     else:
-        msg = f"Found {test_config_labels} on the PR so only these test configs are run"
-        info(msg)
         # When the filter test matrix contain matches or if a valid test config label
         # is found in the PR, return the filtered test matrix
         return filtered_test_matrix
 
 
-def filter_selected_test_configs(
-    test_matrix: Dict[str, List[Any]], selected_test_configs: Set[str]
-) -> Dict[str, List[Any]]:
-    """
-    Keep only the selected configs if the list if not empty. Otherwise, keep all test configs.
-    This filter is used when the workflow is dispatched manually.
-    """
-    if not selected_test_configs:
-        return test_matrix
-
-    filtered_test_matrix: Dict[str, List[Any]] = {"include": []}
-    for entry in test_matrix.get("include", []):
-        config_name = entry.get("config", "")
-        if not config_name:
-            continue
-
-        if config_name in selected_test_configs:
-            filtered_test_matrix["include"].append(entry)
-
-    return filtered_test_matrix
-
-
 def set_periodic_modes(
     test_matrix: Dict[str, List[Any]], job_name: Optional[str]
 ) -> Dict[str, List[Any]]:
@@ -386,33 +374,30 @@ def process_jobs(
         # - If the target record has the job (config) name, only that test config
         #   will be skipped or marked as unstable
         if not target_job_cfg:
-            msg = (
+            print(
                 f"Issue {target_url} created by {author} has {issue_type.value} "
                 + f"all CI jobs for {workflow} / {job_name}"
             )
-            info(msg)
             return _filter_jobs(
                 test_matrix=test_matrix,
                 issue_type=issue_type,
             )
 
         if target_job_cfg == BUILD_JOB_NAME:
-            msg = (
+            print(
                 f"Issue {target_url} created by {author} has {issue_type.value} "
                 + f"the build job for {workflow} / {job_name}"
             )
-            info(msg)
             return _filter_jobs(
                 test_matrix=test_matrix,
                 issue_type=issue_type,
             )
 
         if target_job_cfg in (TEST_JOB_NAME, BUILD_AND_TEST_JOB_NAME):
-            msg = (
+            print(
                 f"Issue {target_url} created by {author} has {issue_type.value} "
                 + f"all the test jobs for {workflow} / {job_name}"
             )
-            info(msg)
             return _filter_jobs(
                 test_matrix=test_matrix,
                 issue_type=issue_type,
@@ -478,7 +463,7 @@ def parse_reenabled_issues(s: Optional[str]) -> List[str]:
 
 
 def get_reenabled_issues(pr_body: str = "") -> List[str]:
-    default_branch = f"origin/{os.environ.get('GIT_DEFAULT_BRANCH', 'main')}"
+    default_branch = os.getenv("GIT_DEFAULT_BRANCH", "main")
     try:
         commit_messages = subprocess.check_output(
             f"git cherry -v {default_branch}".split(" ")
@@ -509,15 +494,10 @@ def perform_misc_tasks(
         "ci-no-test-timeout", check_for_setting(labels, pr_body, "ci-no-test-timeout")
     )
     set_output("ci-no-td", check_for_setting(labels, pr_body, "ci-no-td"))
-    # Only relevant for the one linux distributed cuda job, delete this when TD
-    # is rolled out completely
-    set_output(
-        "ci-td-distributed", check_for_setting(labels, pr_body, "ci-td-distributed")
-    )
 
     # Obviously, if the job name includes unstable, then this is an unstable job
     is_unstable = job_name and IssueType.UNSTABLE.value in job_name
-    if not is_unstable and test_matrix and test_matrix.get("include"):
+    if not is_unstable and test_matrix:
         # Even when the job name doesn't mention unstable, we will also mark it as
         # unstable when the test matrix only includes unstable jobs. Basically, this
         # logic allows build or build-and-test jobs to be marked as unstable too.
@@ -587,16 +567,6 @@ def main() -> None:
         # No PR number, no tag, we can just return the test matrix as it is
         filtered_test_matrix = test_matrix
 
-    if args.selected_test_configs:
-        selected_test_configs = {
-            v.strip().lower()
-            for v in args.selected_test_configs.split(",")
-            if v.strip()
-        }
-        filtered_test_matrix = filter_selected_test_configs(
-            filtered_test_matrix, selected_test_configs
-        )
-
     if args.event_name == "schedule" and args.schedule == "29 8 * * *":
         # we don't want to run the mem leak check or disabled tests on normal
         # periodically scheduled jobs, only the ones at this time

.github/scripts/generate_binary_build_matrix.py

@@ -8,25 +8,22 @@ architectures:
     * CPU
     * Latest CUDA
     * Latest ROCM
-    * Latest XPU
 """
 
 import os
 from typing import Dict, List, Optional, Tuple
 
-
-CUDA_ARCHES = ["11.8", "12.1", "12.4"]
+CUDA_ARCHES = ["11.8", "12.1"]
 
 
-CUDA_ARCHES_FULL_VERSION = {"11.8": "11.8.0", "12.1": "12.1.1", "12.4": "12.4.0"}
+CUDA_ARCHES_FULL_VERSION = {"11.8": "11.8.0", "12.1": "12.1.1"}
 
 
-CUDA_ARCHES_CUDNN_VERSION = {"11.8": "9", "12.1": "9", "12.4": "9"}
+CUDA_ARCHES_CUDNN_VERSION = {"11.8": "8", "12.1": "8"}
 
 
-ROCM_ARCHES = ["6.0", "6.1"]
+ROCM_ARCHES = ["5.7", "6.0"]
 
-XPU_ARCHES = ["xpu"]
 
 CPU_CXX11_ABI_ARCH = ["cpu-cxx11-abi"]
 
@@ -34,53 +31,33 @@ CPU_CXX11_ABI_ARCH = ["cpu-cxx11-abi"]
 CPU_AARCH64_ARCH = ["cpu-aarch64"]
 
 
-CPU_S390X_ARCH = ["cpu-s390x"]
-
-
-CUDA_AARCH64_ARCH = ["cuda-aarch64"]
-
-
 PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {
     "11.8": (
         "nvidia-cuda-nvrtc-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | "  # noqa: B950
         "nvidia-cuda-runtime-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cuda-cupti-cu11==11.8.87; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cudnn-cu11==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cudnn-cu11==8.7.0.84; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cublas-cu11==11.11.3.6; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cufft-cu11==10.9.0.58; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-curand-cu11==10.3.0.86; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu11==11.4.1.48; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu11==11.7.5.86; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nccl-cu11==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-nccl-cu11==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
     "12.1": (
         "nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | "  # noqa: B950
         "nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cudnn-cu12==8.9.2.26; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-nccl-cu12==2.20.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
-    "12.4": (
-        "nvidia-cuda-nvrtc-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cuda-runtime-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cuda-cupti-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cublas-cu12==12.4.2.65; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cufft-cu12==11.2.0.44; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-curand-cu12==10.3.5.119; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusolver-cu12==11.6.0.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusparse-cu12==12.3.0.142; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nvtx-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nvjitlink-cu12==12.4.99; platform_system == 'Linux' and platform_machine == 'x86_64'"
-    ),
 }
 
 
@@ -135,16 +112,10 @@ def arch_type(arch_version: str) -> str:
         return "cuda"
     elif arch_version in ROCM_ARCHES:
         return "rocm"
-    elif arch_version in XPU_ARCHES:
-        return "xpu"
     elif arch_version in CPU_CXX11_ABI_ARCH:
         return "cpu-cxx11-abi"
     elif arch_version in CPU_AARCH64_ARCH:
         return "cpu-aarch64"
-    elif arch_version in CPU_S390X_ARCH:
-        return "cpu-s390x"
-    elif arch_version in CUDA_AARCH64_ARCH:
-        return "cuda-aarch64"
     else:  # arch_version should always be "cpu" in this case
         return "cpu"
 
@@ -161,12 +132,9 @@ WHEEL_CONTAINER_IMAGES = {
         gpu_arch: f"pytorch/manylinux-builder:rocm{gpu_arch}-{DEFAULT_TAG}"
         for gpu_arch in ROCM_ARCHES
     },
-    "xpu": f"pytorch/manylinux2_28-builder:xpu-{DEFAULT_TAG}",
     "cpu": f"pytorch/manylinux-builder:cpu-{DEFAULT_TAG}",
     "cpu-cxx11-abi": f"pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-{DEFAULT_TAG}",
     "cpu-aarch64": f"pytorch/manylinuxaarch64-builder:cpu-aarch64-{DEFAULT_TAG}",
-    "cpu-s390x": f"pytorch/manylinuxs390x-builder:cpu-s390x-{DEFAULT_TAG}",
-    "cuda-aarch64": f"pytorch/manylinuxaarch64-builder:cuda12.4-{DEFAULT_TAG}",
 }
 
 CONDA_CONTAINER_IMAGES = {
@@ -223,11 +191,8 @@ def translate_desired_cuda(gpu_arch_type: str, gpu_arch_version: str) -> str:
         "cpu": "cpu",
         "cpu-aarch64": "cpu",
         "cpu-cxx11-abi": "cpu-cxx11-abi",
-        "cpu-s390x": "cpu",
         "cuda": f"cu{gpu_arch_version.replace('.', '')}",
-        "cuda-aarch64": "cu124",
         "rocm": f"rocm{gpu_arch_version}",
-        "xpu": "xpu",
     }.get(gpu_arch_type, gpu_arch_version)
 
 
@@ -307,11 +272,11 @@ def generate_libtorch_matrix(
                     "libtorch_variant": libtorch_variant,
                     "libtorch_config": abi_version if os == "windows" else "",
                     "devtoolset": abi_version if os != "windows" else "",
-                    "container_image": (
-                        LIBTORCH_CONTAINER_IMAGES[(arch_version, abi_version)]
-                        if os != "windows"
-                        else ""
-                    ),
+                    "container_image": LIBTORCH_CONTAINER_IMAGES[
+                        (arch_version, abi_version)
+                    ]
+                    if os != "windows"
+                    else "",
                     "package_type": "libtorch",
                     "build_name": f"libtorch-{gpu_arch_type}{gpu_arch_version}-{libtorch_variant}-{abi_version}".replace(
                         ".", "_"
@@ -327,28 +292,24 @@ def generate_wheels_matrix(
     python_versions: Optional[List[str]] = None,
 ) -> List[Dict[str, str]]:
     package_type = "wheel"
-    if os == "linux" or os == "linux-aarch64" or os == "linux-s390x":
-        # NOTE: We only build manywheel packages for x86_64 and aarch64 and s390x linux
+    if os == "linux" or os == "linux-aarch64":
+        # NOTE: We only build manywheel packages for x86_64 and aarch64 linux
         package_type = "manywheel"
 
     if python_versions is None:
-        python_versions = FULL_PYTHON_VERSIONS + ["3.13"]
+        python_versions = FULL_PYTHON_VERSIONS
 
     if arches is None:
         # Define default compute archivectures
         arches = ["cpu"]
         if os == "linux":
-            arches += CPU_CXX11_ABI_ARCH + CUDA_ARCHES + ROCM_ARCHES + XPU_ARCHES
+            arches += CPU_CXX11_ABI_ARCH + CUDA_ARCHES + ROCM_ARCHES
         elif os == "windows":
             arches += CUDA_ARCHES
         elif os == "linux-aarch64":
             # Only want the one arch as the CPU type is different and
             # uses different build/test scripts
-            arches = ["cpu-aarch64", "cuda-aarch64"]
-        elif os == "linux-s390x":
-            # Only want the one arch as the CPU type is different and
-            # uses different build/test scripts
-            arches = ["cpu-s390x"]
+            arches = ["cpu-aarch64"]
 
     ret: List[Dict[str, str]] = []
     for python_version in python_versions:
@@ -359,24 +320,11 @@ def generate_wheels_matrix(
                 if arch_version == "cpu"
                 or arch_version == "cpu-cxx11-abi"
                 or arch_version == "cpu-aarch64"
-                or arch_version == "cpu-s390x"
-                or arch_version == "cuda-aarch64"
-                or arch_version == "xpu"
                 else arch_version
             )
 
-            # TODO: Enable python 3.13 on rocm, xpu, aarch64, windows
-            if (
-                gpu_arch_type in ["rocm", "xpu"] or os != "linux"
-            ) and python_version == "3.13":
-                continue
-
             # 12.1 linux wheels require PYTORCH_EXTRA_INSTALL_REQUIREMENTS to install
-            if (
-                arch_version in ["12.4", "12.1", "11.8"]
-                and os == "linux"
-                or arch_version == "cuda-aarch64"
-            ):
+            if arch_version in ["12.1", "11.8"] and os == "linux":
                 ret.append(
                     {
                         "python_version": python_version,
@@ -385,64 +333,15 @@ def generate_wheels_matrix(
                         "desired_cuda": translate_desired_cuda(
                             gpu_arch_type, gpu_arch_version
                         ),
-                        "devtoolset": (
-                            "cxx11-abi" if arch_version == "cuda-aarch64" else ""
-                        ),
+                        "devtoolset": "",
                         "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
                         "package_type": package_type,
-                        "pytorch_extra_install_requirements": (
-                            PYTORCH_EXTRA_INSTALL_REQUIREMENTS[arch_version]  # fmt: skip
-                            if os != "linux-aarch64"
-                            else ""
-                        ),
+                        "pytorch_extra_install_requirements": PYTORCH_EXTRA_INSTALL_REQUIREMENTS[arch_version],  # fmt: skip
                         "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}".replace(  # noqa: B950
                             ".", "_"
                         ),
                     }
                 )
-                if arch_version != "cuda-aarch64":
-                    ret.append(
-                        {
-                            "python_version": python_version,
-                            "gpu_arch_type": gpu_arch_type,
-                            "gpu_arch_version": gpu_arch_version,
-                            "desired_cuda": translate_desired_cuda(
-                                gpu_arch_type, gpu_arch_version
-                            ),
-                            "use_split_build": "True",
-                            "devtoolset": "",
-                            "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
-                            "package_type": package_type,
-                            "pytorch_extra_install_requirements": (
-                                PYTORCH_EXTRA_INSTALL_REQUIREMENTS[arch_version]  # fmt: skip
-                                if os != "linux-aarch64"
-                                else ""
-                            ),
-                            "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}-split".replace(  # noqa: B950
-                                ".", "_"
-                            ),
-                        }
-                    )
-                    # Special build building to use on Colab. PyThon 3.10 for 12.1 CUDA
-                    if python_version == "3.10" and arch_version == "12.1":
-                        ret.append(
-                            {
-                                "python_version": python_version,
-                                "gpu_arch_type": gpu_arch_type,
-                                "gpu_arch_version": gpu_arch_version,
-                                "desired_cuda": translate_desired_cuda(
-                                    gpu_arch_type, gpu_arch_version
-                                ),
-                                "use_split_build": "False",
-                                "devtoolset": "",
-                                "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
-                                "package_type": package_type,
-                                "pytorch_extra_install_requirements": "",
-                                "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}-full".replace(  # noqa: B950
-                                    ".", "_"
-                                ),
-                            }
-                        )
             else:
                 ret.append(
                     {
@@ -452,26 +351,21 @@ def generate_wheels_matrix(
                         "desired_cuda": translate_desired_cuda(
                             gpu_arch_type, gpu_arch_version
                         ),
-                        "devtoolset": (
-                            "cxx11-abi"
-                            if arch_version in ["cpu-cxx11-abi", "xpu"]
-                            else ""
-                        ),
+                        "devtoolset": "cxx11-abi"
+                        if arch_version == "cpu-cxx11-abi"
+                        else "",
                         "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
                         "package_type": package_type,
                         "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}".replace(
                             ".", "_"
                         ),
-                        "pytorch_extra_install_requirements": (
-                            PYTORCH_EXTRA_INSTALL_REQUIREMENTS["12.1"]  # fmt: skip
-                            if os != "linux"
-                            else ""
-                        ),
+                        "pytorch_extra_install_requirements":
+                        PYTORCH_EXTRA_INSTALL_REQUIREMENTS["12.1"]  # fmt: skip
+                        if os != "linux" else "",
                     }
                 )
     return ret
 
 
-validate_nccl_dep_consistency("12.4")
 validate_nccl_dep_consistency("12.1")
 validate_nccl_dep_consistency("11.8")