[prototype] Invoke subgraph higher order op

.buckconfig.oss

@@ -21,3 +21,6 @@
   cxx = /usr/bin/clang++
   cxxpp = /usr/bin/clang++
   ld = /usr/bin/clang++
+
+[project]
+  default_flavors_mode=all

.ci/docker/build.sh

@@ -286,23 +286,23 @@ case "$image" in
     TRITON=yes
     ;;
   pytorch-linux-focal-rocm-n-1-py3)
-    ANACONDA_PYTHON_VERSION=3.10
+    ANACONDA_PYTHON_VERSION=3.8
     GCC_VERSION=9
     PROTOBUF=yes
     DB=yes
     VISION=yes
-    ROCM_VERSION=6.1
+    ROCM_VERSION=6.0
     NINJA_VERSION=1.9.0
     CONDA_CMAKE=yes
     TRITON=yes
     ;;
   pytorch-linux-focal-rocm-n-py3)
-    ANACONDA_PYTHON_VERSION=3.10
+    ANACONDA_PYTHON_VERSION=3.8
     GCC_VERSION=9
     PROTOBUF=yes
     DB=yes
     VISION=yes
-    ROCM_VERSION=6.2
+    ROCM_VERSION=6.1
     NINJA_VERSION=1.9.0
     CONDA_CMAKE=yes
     TRITON=yes
@@ -355,12 +355,6 @@ case "$image" in
     CONDA_CMAKE=yes
     VISION=yes
     ;;
-  pytorch-linux-jammy-py3-clang18-asan)
-    ANACONDA_PYTHON_VERSION=3.10
-    CLANG_VERSION=18
-    CONDA_CMAKE=yes
-    VISION=yes
-    ;;
   pytorch-linux-jammy-py3.9-gcc11)
     ANACONDA_PYTHON_VERSION=3.9
     GCC_VERSION=11
@@ -385,14 +379,6 @@ case "$image" in
     GCC_VERSION=11
     CONDA_CMAKE=yes
     HALIDE=yes
-    TRITON=yes
-    ;;
-  pytorch-linux-jammy-py3.12-triton-cpu)
-    CUDA_VERSION=12.4
-    ANACONDA_PYTHON_VERSION=3.12
-    GCC_VERSION=11
-    CONDA_CMAKE=yes
-    TRITON_CPU=yes
     ;;
   pytorch-linux-focal-linter)
     # TODO: Use 3.9 here because of this issue https://github.com/python/mypy/issues/13627.
@@ -523,7 +509,6 @@ docker build \
        --build-arg "UCC_COMMIT=${UCC_COMMIT}" \
        --build-arg "CONDA_CMAKE=${CONDA_CMAKE}" \
        --build-arg "TRITON=${TRITON}" \
-       --build-arg "TRITON_CPU=${TRITON_CPU}" \
        --build-arg "ONNX=${ONNX}" \
        --build-arg "DOCS=${DOCS}" \
        --build-arg "INDUCTOR_BENCHMARKS=${INDUCTOR_BENCHMARKS}" \

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

@@ -1 +0,0 @@
-c7711371cace304afe265c1ffa906415ab82fc66

.ci/docker/ci_commit_pins/triton.txt

@@ -1 +1 @@
-cf34004b8a67d290a962da166f5aa2fc66751326
+5fe38ffd73c2ac6ed6323b554205186696631c6f

.ci/docker/common/install_clang.sh

@@ -13,18 +13,11 @@ if [ -n "$CLANG_VERSION" ]; then
   elif [[ $UBUNTU_VERSION == 22.04 ]]; then
     # work around ubuntu apt-get conflicts
     sudo apt-get -y -f install
-    wget --no-check-certificate -O - https://apt.llvm.org/llvm-snapshot.gpg.key | sudo apt-key add  -
-    if [[ $CLANG_VERSION == 18 ]]; then
-      apt-add-repository "deb http://apt.llvm.org/jammy/ llvm-toolchain-jammy-18 main"
-    fi
   fi
 
   sudo apt-get update
-  if [[ $CLANG_VERSION -ge 18 ]]; then
-    apt-get install -y libomp-${CLANG_VERSION}-dev libclang-rt-${CLANG_VERSION}-dev clang-"$CLANG_VERSION" llvm-"$CLANG_VERSION"
-  else
-    apt-get install -y --no-install-recommends clang-"$CLANG_VERSION" llvm-"$CLANG_VERSION"
-  fi
+  apt-get install -y --no-install-recommends clang-"$CLANG_VERSION"
+  apt-get install -y --no-install-recommends llvm-"$CLANG_VERSION"
 
   # Install dev version of LLVM.
   if [ -n "$LLVMDEV" ]; then

.ci/docker/common/install_conda.sh

@@ -65,10 +65,23 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
 
   # Install PyTorch conda deps, as per https://github.com/pytorch/pytorch README
   if [[ $(uname -m) == "aarch64" ]]; then
-    conda_install "openblas==0.3.25=*openmp*"
+    CONDA_COMMON_DEPS="astunparse pyyaml setuptools openblas==0.3.25=*openmp* ninja==1.11.1 scons==4.5.2"
+
+    if [ "$ANACONDA_PYTHON_VERSION" = "3.8" ]; then
+      NUMPY_VERSION=1.24.4
+    else
+      NUMPY_VERSION=1.26.2
+    fi
   else
-    conda_install "mkl=2021.4.0 mkl-include=2021.4.0"
+    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
+      NUMPY_VERSION=1.26.0
+    else
+      NUMPY_VERSION=1.21.2
+    fi
   fi
+  conda_install ${CONDA_COMMON_DEPS}
 
   # Install llvm-8 as it is required to compile llvmlite-0.30.0 from source
   # and libpython-static for torch deploy
@@ -90,6 +103,8 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
 
   # Install some other packages, including those needed for Python test reporting
   pip_install -r /opt/conda/requirements-ci.txt
+  pip_install numpy=="$NUMPY_VERSION"
+  pip_install -U scikit-learn
 
   if [ -n "$DOCS" ]; then
     apt-get update

.ci/docker/common/install_cpython.sh

@@ -7,7 +7,7 @@ PYTHON_DOWNLOAD_GITHUB_BRANCH=https://github.com/python/cpython/archive/refs/hea
 GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py
 
 # Python versions to be installed in /opt/$VERSION_NO
-CPYTHON_VERSIONS=${CPYTHON_VERSIONS:-"3.8.1 3.9.0 3.10.1 3.11.0 3.12.0 3.13.0 3.13.0t"}
+CPYTHON_VERSIONS=${CPYTHON_VERSIONS:-"3.8.1 3.9.0 3.10.1 3.11.0 3.12.0 3.13.0"}
 
 function check_var {
     if [ -z "$1" ]; then
@@ -22,13 +22,6 @@ function do_cpython_build {
     check_var $py_ver
     check_var $py_folder
     tar -xzf Python-$py_ver.tgz
-
-    local additional_flags=""
-    if [ "$py_ver" == "3.13.0t" ]; then
-        additional_flags=" --disable-gil"
-        mv cpython-3.13/ cpython-3.13t/
-    fi
-
     pushd $py_folder
 
     local prefix="/opt/_internal/cpython-${py_ver}"
@@ -44,10 +37,8 @@ function do_cpython_build {
         local openssl_flags="--with-openssl=${WITH_OPENSSL} --with-openssl-rpath=auto"
     fi
 
-
-
     # -Wformat added for https://bugs.python.org/issue17547 on Python 2.6
-    CFLAGS="-Wformat" ./configure --prefix=${prefix} ${openssl_flags} ${shared_flags} ${additional_flags} > /dev/null
+    CFLAGS="-Wformat" ./configure --prefix=${prefix} ${openssl_flags} ${shared_flags} > /dev/null
 
     make -j40 > /dev/null
     make install > /dev/null
@@ -78,14 +69,7 @@ function build_cpython {
     check_var $py_ver
     check_var $PYTHON_DOWNLOAD_URL
     local py_ver_folder=$py_ver
-
-    if [ "$py_ver" = "3.13.0t" ]; then
-        PY_VER_SHORT="3.13"
-        PYT_VER_SHORT="3.13t"
-        check_var $PYTHON_DOWNLOAD_GITHUB_BRANCH
-        wget $PYTHON_DOWNLOAD_GITHUB_BRANCH/$PY_VER_SHORT.tar.gz -O Python-$py_ver.tgz
-        do_cpython_build $py_ver cpython-$PYT_VER_SHORT
-    elif [ "$py_ver" = "3.13.0" ]; then
+    if [ "$py_ver" = "3.13.0" ]; then
         PY_VER_SHORT="3.13"
         check_var $PYTHON_DOWNLOAD_GITHUB_BRANCH
         wget $PYTHON_DOWNLOAD_GITHUB_BRANCH/$PY_VER_SHORT.tar.gz -O Python-$py_ver.tgz

.ci/docker/common/install_cuda.sh

@@ -105,7 +105,7 @@ function install_121 {
 }
 
 function install_124 {
-  echo "Installing CUDA 12.4.1 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.2"
+  echo "Installing CUDA 12.4.1 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.5.2"
   rm -rf /usr/local/cuda-12.4 /usr/local/cuda
   # install CUDA 12.4.1 in the same container
   wget -q https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda_12.4.1_550.54.15_linux.run

.ci/docker/common/install_cuda_aarch64.sh

@@ -5,19 +5,19 @@ set -ex
 
 NCCL_VERSION=v2.21.5-1
 
-function install_cusparselt_062 {
+function install_cusparselt_052 {
     # cuSparseLt license: https://docs.nvidia.com/cuda/cusparselt/license.html
     mkdir tmp_cusparselt && pushd tmp_cusparselt
-    wget -q https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-sbsa/libcusparse_lt-linux-sbsa-0.6.2.3-archive.tar.xz
-    tar xf libcusparse_lt-linux-sbsa-0.6.2.3-archive.tar.xz
-    cp -a libcusparse_lt-linux-sbsa-0.6.2.3-archive/include/* /usr/local/cuda/include/
-    cp -a libcusparse_lt-linux-sbsa-0.6.2.3-archive/lib/* /usr/local/cuda/lib64/
+    wget -q https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-sbsa/libcusparse_lt-linux-sbsa-0.5.2.1-archive.tar.xz
+    tar xf libcusparse_lt-linux-sbsa-0.5.2.1-archive.tar.xz
+    cp -a libcusparse_lt-linux-sbsa-0.5.2.1-archive/include/* /usr/local/cuda/include/
+    cp -a libcusparse_lt-linux-sbsa-0.5.2.1-archive/lib/* /usr/local/cuda/lib64/
     popd
     rm -rf tmp_cusparselt
 }
 
 function install_124 {
-  echo "Installing CUDA 12.4.1 and cuDNN 9.1 and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.2"
+  echo "Installing CUDA 12.4.1 and cuDNN 9.1 and NCCL ${NCCL_VERSION} and cuSparseLt-0.5.2"
   rm -rf /usr/local/cuda-12.4 /usr/local/cuda
   # install CUDA 12.4.1 in the same container
   wget -q https://developer.download.nvidia.com/compute/cuda/12.4.1/local_installers/cuda_12.4.1_550.54.15_linux_sbsa.run
@@ -44,7 +44,7 @@ function install_124 {
   cd ..
   rm -rf nccl
 
-  install_cusparselt_062
+  install_cusparselt_052
 
   ldconfig
 }

.ci/docker/common/install_cusparselt.sh

@@ -5,7 +5,7 @@ set -ex
 # cuSPARSELt license: https://docs.nvidia.com/cuda/cusparselt/license.html
 mkdir tmp_cusparselt && cd tmp_cusparselt
 
-if [[ ${CUDA_VERSION:0:4} =~ ^12\.[2-6]$ ]]; then
+if [[ ${CUDA_VERSION:0:4} =~ ^12\.[2-4]$ ]]; then
     arch_path='sbsa'
     export TARGETARCH=${TARGETARCH:-$(uname -m)}
     if [ ${TARGETARCH} = 'amd64' ] || [ "${TARGETARCH}" = 'x86_64' ]; then

.ci/docker/common/install_miopen.sh

@@ -10,21 +10,6 @@ if [[ -z $ROCM_VERSION ]]; then
     exit 1;
 fi
 
-IS_UBUNTU=0
-ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
-case "$ID" in
-  ubuntu)
-    IS_UBUNTU=1
-    ;;
-  centos)
-    IS_UBUNTU=0
-    ;;
-  *)
-    echo "Unable to determine OS..."
-    exit 1
-    ;;
-esac
-
 # To make version comparison easier, create an integer representation.
 save_IFS="$IFS"
 IFS=. ROCM_VERSION_ARRAY=(${ROCM_VERSION})
@@ -72,11 +57,9 @@ MIOPEN_CMAKE_COMMON_FLAGS="
 -DMIOPEN_BUILD_DRIVER=OFF
 "
 # Pull MIOpen repo and set DMIOPEN_EMBED_DB based on ROCm version
-if [[ $ROCM_INT -ge 60300 ]]; then
-    echo "ROCm 6.3+ MIOpen does not need any patches, do not build from source"
+if [[ $ROCM_INT -ge 60200 ]] && [[ $ROCM_INT -lt 60300 ]]; then
+    echo "ROCm 6.2 MIOpen does not need any patches, do not build from source"
     exit 0
-elif [[ $ROCM_INT -ge 60200 ]] && [[ $ROCM_INT -lt 60300 ]]; then
-    MIOPEN_BRANCH="release/rocm-rel-6.2-staging"
 elif [[ $ROCM_INT -ge 60100 ]] && [[ $ROCM_INT -lt 60200 ]]; then
     echo "ROCm 6.1 MIOpen does not need any patches, do not build from source"
     exit 0
@@ -110,21 +93,12 @@ else
     exit 1
 fi
 
-
-if [[ ${IS_UBUNTU} == 1 ]]; then
-  apt-get remove -y miopen-hip
-else
-  yum remove -y miopen-hip
-fi
+yum remove -y miopen-hip
 
 git clone https://github.com/ROCm/MIOpen -b ${MIOPEN_BRANCH}
 pushd MIOpen
 # remove .git to save disk space since CI runner was running out
 rm -rf .git
-# Don't build CK to save docker build time
-if [[ $ROCM_INT -ge 60200 ]]; then
-    sed -i '/composable_kernel/d' requirements.txt
-fi
 # Don't build MLIR to save docker build time
 # since we are disabling MLIR backend for MIOpen anyway
 if [[ $ROCM_INT -ge 50400 ]] && [[ $ROCM_INT -lt 50500 ]]; then
@@ -137,15 +111,10 @@ cmake -P install_deps.cmake --minimum
 
 # clean up since CI runner was running out of disk space
 rm -rf /tmp/*
-if [[ ${IS_UBUNTU} == 1 ]]; then
-  apt-get autoclean && apt-get clean
-  rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
-else
-  yum clean all
-  rm -rf /var/cache/yum
-  rm -rf /var/lib/yum/yumdb
-  rm -rf /var/lib/yum/history
-fi
+yum clean all
+rm -rf /var/cache/yum
+rm -rf /var/lib/yum/yumdb
+rm -rf /var/lib/yum/history
 
 ## Build MIOpen
 mkdir -p build
@@ -162,11 +131,7 @@ make -j $(nproc) package
 # clean up since CI runner was running out of disk space
 rm -rf /usr/local/cget
 
-if [[ ${IS_UBUNTU} == 1 ]]; then
-  sudo dpkg -i miopen-hip*.deb
-else
-  yum install -y miopen-*.rpm
-fi
+yum install -y miopen-*.rpm
 
 popd
 rm -rf MIOpen

.ci/docker/common/install_onnx.sh

@@ -32,7 +32,7 @@ pip_install coloredlogs packaging
 
 pip_install onnxruntime==1.18.1
 pip_install onnx==1.16.2
-pip_install onnxscript==0.1.0.dev20241009 --no-deps
+pip_install onnxscript==0.1.0.dev20240831 --no-deps
 # required by onnxscript
 pip_install ml_dtypes
 

.ci/docker/common/install_triton.sh

@@ -15,11 +15,8 @@ conda_reinstall() {
 if [ -n "${XPU_VERSION}" ]; then
   TRITON_REPO="https://github.com/intel/intel-xpu-backend-for-triton"
   TRITON_TEXT_FILE="triton-xpu"
-elif [ -n "${TRITON_CPU}" ]; then
-  TRITON_REPO="https://github.com/triton-lang/triton-cpu"
-  TRITON_TEXT_FILE="triton-cpu"
 else
-  TRITON_REPO="https://github.com/triton-lang/triton"
+  TRITON_REPO="https://github.com/openai/triton"
   TRITON_TEXT_FILE="triton"
 fi
 
@@ -47,10 +44,9 @@ chown -R jenkins /var/lib/jenkins/triton
 chgrp -R jenkins /var/lib/jenkins/triton
 pushd /var/lib/jenkins/
 
-as_jenkins git clone --recursive ${TRITON_REPO} triton
+as_jenkins git clone ${TRITON_REPO} triton
 cd triton
 as_jenkins git checkout ${TRITON_PINNED_COMMIT}
-as_jenkins git submodule update --init --recursive
 cd python
 
 # TODO: remove patch setup.py once we have a proper fix for https://github.com/triton-lang/triton/issues/4527

.ci/docker/conda/build.sh

@@ -37,12 +37,6 @@ esac
 
 (
   set -x
-  # TODO: Remove LimitNOFILE=1048576 patch once https://github.com/pytorch/test-infra/issues/5712
-  # is resolved. This patch is required in order to fix timing out of Docker build on Amazon Linux 2023.
-  sudo sed -i s/LimitNOFILE=infinity/LimitNOFILE=1048576/ /usr/lib/systemd/system/docker.service
-  sudo systemctl daemon-reload
-  sudo systemctl restart docker
-
   docker build \
     --target final \
     --progress plain \

.ci/docker/manywheel/Dockerfile

@@ -10,7 +10,6 @@ ENV LANG en_US.UTF-8
 ENV LANGUAGE en_US.UTF-8
 
 ARG DEVTOOLSET_VERSION=9
-
 # Note: This is required patch since CentOS have reached EOL
 # otherwise any yum install setp will fail
 RUN sed -i s/mirror.centos.org/vault.centos.org/g /etc/yum.repos.d/*.repo

.ci/docker/manywheel/build.sh

@@ -124,14 +124,7 @@ if [[ -n ${MANY_LINUX_VERSION} && -z ${DOCKERFILE_SUFFIX} ]]; then
 fi
 (
     set -x
-
-    # TODO: Remove LimitNOFILE=1048576 patch once https://github.com/pytorch/test-infra/issues/5712
-    # is resolved. This patch is required in order to fix timing out of Docker build on Amazon Linux 2023.
-    sudo sed -i s/LimitNOFILE=infinity/LimitNOFILE=1048576/ /usr/lib/systemd/system/docker.service
-    sudo systemctl daemon-reload
-    sudo systemctl restart docker
-
-    DOCKER_BUILDKIT=1 docker build  \
+    DOCKER_BUILDKIT=1 docker build \
         ${DOCKER_GPU_BUILD_ARG} \
         --build-arg "GPU_IMAGE=${GPU_IMAGE}" \
         --target "${TARGET}" \

.ci/docker/requirements-ci.txt

@@ -90,7 +90,7 @@ librosa>=0.6.2 ; python_version < "3.11"
 #Pinned versions:
 #test that import:
 
-mypy==1.11.2
+mypy==1.10.0
 # Pin MyPy version because new errors are likely to appear with each release
 #Description: linter
 #Pinned versions: 1.10.0
@@ -118,7 +118,7 @@ numba==0.55.2 ; python_version == "3.10"
 
 #numpy
 #Description: Provides N-dimensional arrays and linear algebra
-#Pinned versions: 1.26.2
+#Pinned versions: 1.20
 #test that import: test_view_ops.py, test_unary_ufuncs.py, test_type_promotion.py,
 #test_type_info.py, test_torch.py, test_tensorexpr_pybind.py, test_tensorexpr.py,
 #test_tensorboard.py, test_tensor_creation_ops.py, test_static_runtime.py,
@@ -128,10 +128,6 @@ numba==0.55.2 ; python_version == "3.10"
 #test_nn.py, test_namedtensor.py, test_linalg.py, test_jit_cuda_fuser.py,
 #test_jit.py, test_indexing.py, test_datapipe.py, test_dataloader.py,
 #test_binary_ufuncs.py
-numpy==1.21.2; python_version == "3.9"
-numpy==1.22.4; python_version == "3.10"
-numpy==1.26.2; python_version == "3.11" or python_version == "3.12"
-numpy==2.1.2; python_version >= "3.13"
 
 #onnxruntime
 #Description: scoring engine for Open Neural Network Exchange (ONNX) models
@@ -143,9 +139,9 @@ opt-einsum==3.3
 #Pinned versions: 3.3
 #test that import: test_linalg.py
 
-optree==0.13.0
+optree==0.12.1
 #Description: A library for tree manipulation
-#Pinned versions: 0.13.0
+#Pinned versions: 0.12.1
 #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,
@@ -326,6 +322,7 @@ lxml==5.0.0
 
 PyGithub==2.3.0
 
+sympy==1.12.1 ; python_version == "3.8"
 sympy==1.13.1 ; python_version >= "3.9"
 #Description: Required by coremltools, also pinned in .github/requirements/pip-requirements-macOS.txt
 #Pinned versions:
@@ -345,26 +342,3 @@ parameterized==0.8.1
 #Description: Parameterizes unittests, both the tests themselves and the entire testing class
 #Pinned versions:
 #test that import:
-
-#Description: required for testing torch/distributed/_tools/sac_estimator.py
-#Pinned versions: 1.24.0
-#test that import: test_sac_estimator.py
-
-pwlf==2.2.1 ; python_version >= "3.8"
-#Description: required for testing torch/distributed/_tools/sac_estimator.py
-#Pinned versions: 2.2.1
-#test that import: test_sac_estimator.py
-
-
-# To build PyTorch itself
-astunparse
-PyYAML
-setuptools
-
-ninja==1.11.1 ; platform_machine == "aarch64"
-scons==4.5.2 ; platform_machine == "aarch64"
-
-pulp==2.9.0 ; python_version >= "3.8"
-#Description: required for testing ilp formulaiton under torch/distributed/_tools
-#Pinned versions: 2.9.0
-#test that import: test_sac_ilp.py

.ci/docker/ubuntu-rocm/Dockerfile

@@ -68,8 +68,6 @@ 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
-ADD ./common/install_miopen.sh install_miopen.sh
-RUN bash ./install_miopen.sh ${ROCM_VERSION} && rm install_miopen.sh
 ENV ROCM_PATH /opt/rocm
 ENV PATH /opt/rocm/bin:$PATH
 ENV PATH /opt/rocm/hcc/bin:$PATH
@@ -123,8 +121,5 @@ RUN bash ./install_cache.sh && rm install_cache.sh
 ARG BUILD_ENVIRONMENT
 ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
 
-# Install LLVM dev version (Defined in the pytorch/builder github repository)
-COPY --from=pytorch/llvm:9.0.1 /opt/llvm /opt/llvm
-
 USER jenkins
 CMD ["bash"]

.ci/docker/ubuntu/Dockerfile

@@ -147,13 +147,6 @@ 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.txt
 
-ARG TRITON_CPU
-COPY ./common/install_triton.sh install_triton.sh
-COPY ./common/common_utils.sh common_utils.sh
-COPY ci_commit_pins/triton-cpu.txt triton-cpu.txt
-RUN if [ -n "${TRITON_CPU}" ]; then bash ./install_triton.sh; fi
-RUN rm install_triton.sh common_utils.sh triton-cpu.txt
-
 ARG EXECUTORCH
 # Build and install executorch
 COPY ./common/install_executorch.sh install_executorch.sh

.ci/libtorch/build.sh

@@ -1,10 +0,0 @@
-#!/usr/bin/env bash
-
-# This is mostly just a shim to manywheel/build.sh
-# TODO: Make this a dedicated script to build just libtorch
-
-set -ex
-
-SCRIPTPATH="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
-
-USE_CUSPARSELT=0 BUILD_PYTHONLESS=1 DESIRED_PYTHON="3.9" ${SCRIPTPATH}/../manywheel/build.sh

.ci/manywheel/LICENSE

@@ -1,21 +0,0 @@
-The MIT License (MIT)
-
-Copyright (c) 2016 manylinux
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.

.ci/manywheel/build.sh

@@ -1,25 +0,0 @@
-#!/usr/bin/env bash
-
-set -ex
-
-SCRIPTPATH="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
-
-case "${GPU_ARCH_TYPE:-BLANK}" in
-    BLANK)
-        # Legacy behavior for CircleCI
-        bash "${SCRIPTPATH}/build_cuda.sh"
-        ;;
-    cuda)
-        bash "${SCRIPTPATH}/build_cuda.sh"
-        ;;
-    rocm)
-        bash "${SCRIPTPATH}/build_rocm.sh"
-        ;;
-    cpu | cpu-cxx11-abi | cpu-s390x | xpu)
-        bash "${SCRIPTPATH}/build_cpu.sh"
-        ;;
-    *)
-        echo "Un-recognized GPU_ARCH_TYPE '${GPU_ARCH_TYPE}', exiting..."
-        exit 1
-        ;;
-esac

.ci/manywheel/build_common.sh

@@ -1,505 +0,0 @@
-#!/usr/bin/env bash
-# meant to be called only from the neighboring build.sh and build_cpu.sh scripts
-
-set -ex
-SOURCE_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null && pwd )"
-
-
-# Require only one python installation
-if [[ -z "$DESIRED_PYTHON" ]]; then
-    echo "Need to set DESIRED_PYTHON env variable"
-    exit 1
-fi
-if [[ -n "$BUILD_PYTHONLESS" && -z "$LIBTORCH_VARIANT" ]]; then
-    echo "BUILD_PYTHONLESS is set, so need LIBTORCH_VARIANT to also be set"
-    echo "LIBTORCH_VARIANT should be one of shared-with-deps shared-without-deps static-with-deps static-without-deps"
-    exit 1
-fi
-
-# Function to retry functions that sometimes timeout or have flaky failures
-retry () {
-    $*  || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
-}
-
-# TODO move this into the Docker images
-OS_NAME=$(awk -F= '/^NAME/{print $2}' /etc/os-release)
-if [[ "$OS_NAME" == *"CentOS Linux"* ]]; then
-    retry yum install -q -y zip openssl
-elif [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
-    retry yum install -q -y zip openssl
-elif [[ "$OS_NAME" == *"Red Hat Enterprise Linux"* ]]; then
-    retry dnf install -q -y zip openssl
-elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
-    # TODO: Remove this once nvidia package repos are back online
-    # Comment out nvidia repositories to prevent them from getting apt-get updated, see https://github.com/pytorch/pytorch/issues/74968
-    # shellcheck disable=SC2046
-    sed -i 's/.*nvidia.*/# &/' $(find /etc/apt/ -type f -name "*.list")
-
-    retry apt-get update
-    retry apt-get -y install zip openssl
-fi
-
-# We use the package name to test the package by passing this to 'pip install'
-# This is the env variable that setup.py uses to name the package. Note that
-# pip 'normalizes' the name first by changing all - to _
-if [[ -z "$TORCH_PACKAGE_NAME" ]]; then
-    TORCH_PACKAGE_NAME='torch'
-fi
-
-if [[ -z "$TORCH_NO_PYTHON_PACKAGE_NAME" ]]; then
-    TORCH_NO_PYTHON_PACKAGE_NAME='torch_no_python'
-fi
-
-TORCH_PACKAGE_NAME="$(echo $TORCH_PACKAGE_NAME | tr '-' '_')"
-TORCH_NO_PYTHON_PACKAGE_NAME="$(echo $TORCH_NO_PYTHON_PACKAGE_NAME | tr '-' '_')"
-echo "Expecting the built wheels to all be called '$TORCH_PACKAGE_NAME' or '$TORCH_NO_PYTHON_PACKAGE_NAME'"
-
-# Version: setup.py uses $PYTORCH_BUILD_VERSION.post$PYTORCH_BUILD_NUMBER if
-# PYTORCH_BUILD_NUMBER > 1
-build_version="$PYTORCH_BUILD_VERSION"
-build_number="$PYTORCH_BUILD_NUMBER"
-if [[ -n "$OVERRIDE_PACKAGE_VERSION" ]]; then
-    # This will be the *exact* version, since build_number<1
-    build_version="$OVERRIDE_PACKAGE_VERSION"
-    build_number=0
-fi
-if [[ -z "$build_version" ]]; then
-    build_version=1.0.0
-fi
-if [[ -z "$build_number" ]]; then
-    build_number=1
-fi
-export PYTORCH_BUILD_VERSION=$build_version
-export PYTORCH_BUILD_NUMBER=$build_number
-
-export CMAKE_LIBRARY_PATH="/opt/intel/lib:/lib:$CMAKE_LIBRARY_PATH"
-export CMAKE_INCLUDE_PATH="/opt/intel/include:$CMAKE_INCLUDE_PATH"
-
-if [[ -e /opt/openssl ]]; then
-    export OPENSSL_ROOT_DIR=/opt/openssl
-    export CMAKE_INCLUDE_PATH="/opt/openssl/include":$CMAKE_INCLUDE_PATH
-fi
-
-# If given a python version like 3.6m or 2.7mu, convert this to the format we
-# expect. The binary CI jobs pass in python versions like this; they also only
-# ever pass one python version, so we assume that DESIRED_PYTHON is not a list
-# in this case
-if [[ -n "$DESIRED_PYTHON" && $DESIRED_PYTHON =~ ([0-9].[0-9]+)t ]]; then
-    python_digits="$(echo $DESIRED_PYTHON | tr -cd [:digit:])"
-    py_majmin="${DESIRED_PYTHON}"
-    DESIRED_PYTHON="cp${python_digits}-cp${python_digits}t"
-elif [[ -n "$DESIRED_PYTHON" && "$DESIRED_PYTHON" != cp* ]]; then
-    python_nodot="$(echo $DESIRED_PYTHON | tr -d m.u)"
-    DESIRED_PYTHON="cp${python_nodot}-cp${python_nodot}"
-    if [[ ${python_nodot} -ge 310 ]]; then
-        py_majmin="${DESIRED_PYTHON:2:1}.${DESIRED_PYTHON:3:2}"
-    else
-        py_majmin="${DESIRED_PYTHON:2:1}.${DESIRED_PYTHON:3:1}"
-    fi
-fi
-
-pydir="/opt/python/$DESIRED_PYTHON"
-export PATH="$pydir/bin:$PATH"
-echo "Will build for Python version: ${DESIRED_PYTHON} with ${python_installation}"
-
-mkdir -p /tmp/$WHEELHOUSE_DIR
-
-export PATCHELF_BIN=/usr/local/bin/patchelf
-patchelf_version=$($PATCHELF_BIN --version)
-echo "patchelf version: " $patchelf_version
-if [[ "$patchelf_version" == "patchelf 0.9" ]]; then
-    echo "Your patchelf version is too old. Please use version >= 0.10."
-    exit 1
-fi
-
-########################################################
-# Compile wheels as well as libtorch
-#######################################################
-if [[ -z "$PYTORCH_ROOT" ]]; then
-    echo "Need to set PYTORCH_ROOT env variable"
-    exit 1
-fi
-pushd "$PYTORCH_ROOT"
-python setup.py clean
-retry pip install -qr requirements.txt
-case ${DESIRED_PYTHON} in
-  cp31*)
-    retry pip install -q --pre numpy==2.1.0
-    ;;
-  # Should catch 3.9+
-  *)
-    retry pip install -q --pre numpy==2.0.2
-    ;;
-esac
-
-if [[ "$DESIRED_DEVTOOLSET" == *"cxx11-abi"* ]]; then
-    export _GLIBCXX_USE_CXX11_ABI=1
-else
-    export _GLIBCXX_USE_CXX11_ABI=0
-fi
-
-if [[ "$DESIRED_CUDA" == *"rocm"* ]]; then
-    echo "Calling build_amd.py at $(date)"
-    python tools/amd_build/build_amd.py
-fi
-
-# This value comes from binary_linux_build.sh (and should only be set to true
-# for master / release branches)
-BUILD_DEBUG_INFO=${BUILD_DEBUG_INFO:=0}
-
-if [[ $BUILD_DEBUG_INFO == "1" ]]; then
-    echo "Building wheel and debug info"
-else
-    echo "BUILD_DEBUG_INFO was not set, skipping debug info"
-fi
-
-if [[ "$DISABLE_RCCL" = 1 ]]; then
-    echo "Disabling NCCL/RCCL in pyTorch"
-    USE_RCCL=0
-    USE_NCCL=0
-    USE_KINETO=0
-else
-    USE_RCCL=1
-    USE_NCCL=1
-    USE_KINETO=1
-fi
-
-echo "Calling setup.py bdist at $(date)"
-
-if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-    echo "Calling setup.py bdist_wheel for split build (BUILD_LIBTORCH_WHL)"
-    time EXTRA_CAFFE2_CMAKE_FLAGS=${EXTRA_CAFFE2_CMAKE_FLAGS[@]} \
-    BUILD_LIBTORCH_WHL=1 BUILD_PYTHON_ONLY=0 \
-    BUILD_LIBTORCH_CPU_WITH_DEBUG=$BUILD_DEBUG_INFO \
-    USE_NCCL=${USE_NCCL} USE_RCCL=${USE_RCCL} USE_KINETO=${USE_KINETO} \
-    python setup.py bdist_wheel -d /tmp/$WHEELHOUSE_DIR
-    echo "Finished setup.py bdist_wheel for split build (BUILD_LIBTORCH_WHL)"
-    echo "Calling setup.py bdist_wheel for split build (BUILD_PYTHON_ONLY)"
-    time EXTRA_CAFFE2_CMAKE_FLAGS=${EXTRA_CAFFE2_CMAKE_FLAGS[@]} \
-    BUILD_LIBTORCH_WHL=0 BUILD_PYTHON_ONLY=1 \
-    BUILD_LIBTORCH_CPU_WITH_DEBUG=$BUILD_DEBUG_INFO \
-    USE_NCCL=${USE_NCCL} USE_RCCL=${USE_RCCL} USE_KINETO=${USE_KINETO} \
-    python setup.py bdist_wheel -d /tmp/$WHEELHOUSE_DIR --cmake
-    echo "Finished setup.py bdist_wheel for split build (BUILD_PYTHON_ONLY)"
-else
-    time CMAKE_ARGS=${CMAKE_ARGS[@]} \
-        EXTRA_CAFFE2_CMAKE_FLAGS=${EXTRA_CAFFE2_CMAKE_FLAGS[@]} \
-        BUILD_LIBTORCH_CPU_WITH_DEBUG=$BUILD_DEBUG_INFO \
-        USE_NCCL=${USE_NCCL} USE_RCCL=${USE_RCCL} USE_KINETO=${USE_KINETO} \
-        python setup.py bdist_wheel -d /tmp/$WHEELHOUSE_DIR
-fi
-echo "Finished setup.py bdist at $(date)"
-
-# Build libtorch packages
-if [[ -n "$BUILD_PYTHONLESS" ]]; then
-    # Now build pythonless libtorch
-    # Note - just use whichever python we happen to be on
-    python setup.py clean
-
-    if [[ $LIBTORCH_VARIANT = *"static"* ]]; then
-        STATIC_CMAKE_FLAG="-DTORCH_STATIC=1"
-    fi
-
-    mkdir -p build
-    pushd build
-    echo "Calling tools/build_libtorch.py at $(date)"
-    time CMAKE_ARGS=${CMAKE_ARGS[@]} \
-         EXTRA_CAFFE2_CMAKE_FLAGS="${EXTRA_CAFFE2_CMAKE_FLAGS[@]} $STATIC_CMAKE_FLAG" \
-         python ../tools/build_libtorch.py
-    echo "Finished tools/build_libtorch.py at $(date)"
-    popd
-
-    mkdir -p libtorch/{lib,bin,include,share}
-    cp -r build/build/lib libtorch/
-
-    # for now, the headers for the libtorch package will just be copied in
-    # from one of the wheels (this is from when this script built multiple
-    # wheels at once)
-    ANY_WHEEL=$(ls /tmp/$WHEELHOUSE_DIR/torch*.whl | head -n1)
-    unzip -d any_wheel $ANY_WHEEL
-    if [[ -d any_wheel/torch/include ]]; then
-        cp -r any_wheel/torch/include libtorch/
-    else
-        cp -r any_wheel/torch/lib/include libtorch/
-    fi
-    cp -r any_wheel/torch/share/cmake libtorch/share/
-    rm -rf any_wheel
-
-    echo $PYTORCH_BUILD_VERSION > libtorch/build-version
-    echo "$(pushd $PYTORCH_ROOT && git rev-parse HEAD)" > libtorch/build-hash
-
-    mkdir -p /tmp/$LIBTORCH_HOUSE_DIR
-
-    if [[ "$DESIRED_DEVTOOLSET" == *"cxx11-abi"* ]]; then
-        LIBTORCH_ABI="cxx11-abi-"
-    else
-        LIBTORCH_ABI=
-    fi
-
-    zip -rq /tmp/$LIBTORCH_HOUSE_DIR/libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-$PYTORCH_BUILD_VERSION.zip libtorch
-    cp /tmp/$LIBTORCH_HOUSE_DIR/libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-$PYTORCH_BUILD_VERSION.zip \
-       /tmp/$LIBTORCH_HOUSE_DIR/libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-latest.zip
-fi
-
-popd
-
-#######################################################################
-# ADD DEPENDENCIES INTO THE WHEEL
-#
-# auditwheel repair doesn't work correctly and is buggy
-# so manually do the work of copying dependency libs and patchelfing
-# and fixing RECORDS entries correctly
-######################################################################
-
-fname_with_sha256() {
-    HASH=$(sha256sum $1 | cut -c1-8)
-    DIRNAME=$(dirname $1)
-    BASENAME=$(basename $1)
-    # Do not rename nvrtc-builtins.so as they are dynamically loaded
-    # by libnvrtc.so
-    # Similarly don't mangle libcudnn and libcublas library names
-    if [[ $BASENAME == "libnvrtc-builtins.s"* || $BASENAME == "libcudnn"* || $BASENAME == "libcublas"*  ]]; then
-        echo $1
-    else
-        INITNAME=$(echo $BASENAME | cut -f1 -d".")
-        ENDNAME=$(echo $BASENAME | cut -f 2- -d".")
-        echo "$DIRNAME/$INITNAME-$HASH.$ENDNAME"
-    fi
-}
-
-fname_without_so_number() {
-    LINKNAME=$(echo $1 | sed -e 's/\.so.*/.so/g')
-    echo "$LINKNAME"
-}
-
-make_wheel_record() {
-    FPATH=$1
-    if echo $FPATH | grep RECORD >/dev/null 2>&1; then
-        # if the RECORD file, then
-        echo "$FPATH,,"
-    else
-        HASH=$(openssl dgst -sha256 -binary $FPATH | openssl base64 | sed -e 's/+/-/g' | sed -e 's/\//_/g' | sed -e 's/=//g')
-        FSIZE=$(ls -nl $FPATH | awk '{print $5}')
-        echo "$FPATH,sha256=$HASH,$FSIZE"
-    fi
-}
-
-replace_needed_sofiles() {
-    find $1 -name '*.so*' | while read sofile; do
-        origname=$2
-        patchedname=$3
-        if [[ "$origname" != "$patchedname" ]] || [[ "$DESIRED_CUDA" == *"rocm"* ]]; 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 $sofile
-            fi
-        fi
-    done
-}
-
-echo 'Built this wheel:'
-ls /tmp/$WHEELHOUSE_DIR
-mkdir -p "/$WHEELHOUSE_DIR"
-mv /tmp/$WHEELHOUSE_DIR/torch*linux*.whl /$WHEELHOUSE_DIR/
-
-if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-    mv /tmp/$WHEELHOUSE_DIR/torch_no_python*.whl /$WHEELHOUSE_DIR/ || true
-fi
-
-if [[ -n "$BUILD_PYTHONLESS" ]]; then
-    mkdir -p /$LIBTORCH_HOUSE_DIR
-    mv /tmp/$LIBTORCH_HOUSE_DIR/*.zip /$LIBTORCH_HOUSE_DIR
-    rm -rf /tmp/$LIBTORCH_HOUSE_DIR
-fi
-rm -rf /tmp/$WHEELHOUSE_DIR
-rm -rf /tmp_dir
-mkdir /tmp_dir
-pushd /tmp_dir
-
-for pkg in /$WHEELHOUSE_DIR/torch_no_python*.whl /$WHEELHOUSE_DIR/torch*linux*.whl /$LIBTORCH_HOUSE_DIR/libtorch*.zip; do
-
-    # if the glob didn't match anything
-    if [[ ! -e $pkg ]]; then
-        continue
-    fi
-
-    rm -rf tmp
-    mkdir -p tmp
-    cd tmp
-    cp $pkg .
-
-    unzip -q $(basename $pkg)
-    rm -f $(basename $pkg)
-
-    if [[ -d torch ]]; then
-        PREFIX=torch
-    else
-        PREFIX=libtorch
-    fi
-
-    if [[ $pkg != *"without-deps"* ]]; then
-        # copy over needed dependent .so files over and tag them with their hash
-        patched=()
-        for filepath in "${DEPS_LIST[@]}"; do
-            filename=$(basename $filepath)
-            destpath=$PREFIX/lib/$filename
-            if [[ "$filepath" != "$destpath" ]]; then
-                cp $filepath $destpath
-            fi
-
-            # ROCm workaround for roctracer dlopens
-            if [[ "$DESIRED_CUDA" == *"rocm"* ]]; then
-                patchedpath=$(fname_without_so_number $destpath)
-            # Keep the so number for XPU dependencies
-            elif [[ "$DESIRED_CUDA" == *"xpu"* ]]; then
-                patchedpath=$destpath
-            else
-                patchedpath=$(fname_with_sha256 $destpath)
-            fi
-            patchedname=$(basename $patchedpath)
-            if [[ "$destpath" != "$patchedpath" ]]; then
-                mv $destpath $patchedpath
-            fi
-            patched+=("$patchedname")
-            echo "Copied $filepath to $patchedpath"
-        done
-
-        echo "patching to fix the so names to the hashed names"
-        for ((i=0;i<${#DEPS_LIST[@]};++i)); do
-            replace_needed_sofiles $PREFIX ${DEPS_SONAME[i]} ${patched[i]}
-            # do the same for caffe2, if it exists
-            if [[ -d caffe2 ]]; then
-                replace_needed_sofiles caffe2 ${DEPS_SONAME[i]} ${patched[i]}
-            fi
-        done
-
-        # copy over needed auxiliary files
-        for ((i=0;i<${#DEPS_AUX_SRCLIST[@]};++i)); do
-            srcpath=${DEPS_AUX_SRCLIST[i]}
-            dstpath=$PREFIX/${DEPS_AUX_DSTLIST[i]}
-            mkdir -p $(dirname $dstpath)
-            cp $srcpath $dstpath
-        done
-    fi
-
-    # set RPATH of _C.so and similar to $ORIGIN, $ORIGIN/lib
-    find $PREFIX -maxdepth 1 -type f -name "*.so*" | while read sofile; do
-        echo "Setting rpath of $sofile to ${C_SO_RPATH:-'$ORIGIN:$ORIGIN/lib'}"
-        $PATCHELF_BIN --set-rpath ${C_SO_RPATH:-'$ORIGIN:$ORIGIN/lib'} ${FORCE_RPATH:-} $sofile
-        $PATCHELF_BIN --print-rpath $sofile
-    done
-
-    # set RPATH of lib/ files to $ORIGIN
-    find $PREFIX/lib -maxdepth 1 -type f -name "*.so*" | 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 $sofile
-    done
-
-    # regenerate the RECORD file with new hashes
-    record_file=$(echo $(basename $pkg) | sed -e 's/-cp.*$/.dist-info\/RECORD/g')
-    if [[ -e $record_file ]]; then
-        echo "Generating new record file $record_file"
-        : > "$record_file"
-        # generate records for folders in wheel
-        find * -type f | while read fname; do
-            make_wheel_record "$fname" >>"$record_file"
-        done
-    fi
-
-    if [[ $BUILD_DEBUG_INFO == "1" ]]; then
-        pushd "$PREFIX/lib"
-
-        # Duplicate library into debug lib
-        cp libtorch_cpu.so libtorch_cpu.so.dbg
-
-        # Keep debug symbols on debug lib
-        strip --only-keep-debug libtorch_cpu.so.dbg
-
-        # Remove debug info from release lib
-        strip --strip-debug libtorch_cpu.so
-
-        objcopy libtorch_cpu.so --add-gnu-debuglink=libtorch_cpu.so.dbg
-
-        # Zip up debug info
-        mkdir -p /tmp/debug
-        mv libtorch_cpu.so.dbg /tmp/debug/libtorch_cpu.so.dbg
-        CRC32=$(objcopy --dump-section .gnu_debuglink=>(tail -c4 | od -t x4 -An | xargs echo) libtorch_cpu.so)
-
-        pushd /tmp
-        PKG_NAME=$(basename "$pkg" | sed 's/\.whl$//g')
-        zip /tmp/debug-whl-libtorch-"$PKG_NAME"-"$CRC32".zip /tmp/debug/libtorch_cpu.so.dbg
-        cp /tmp/debug-whl-libtorch-"$PKG_NAME"-"$CRC32".zip "$PYTORCH_FINAL_PACKAGE_DIR"
-        popd
-
-        popd
-    fi
-
-    # zip up the wheel back
-    zip -rq $(basename $pkg) $PREIX*
-
-    # replace original wheel
-    rm -f $pkg
-    mv $(basename $pkg) $pkg
-    cd ..
-    rm -rf tmp
-done
-
-# Copy wheels to host machine for persistence before testing
-if [[ -n "$PYTORCH_FINAL_PACKAGE_DIR" ]]; then
-    mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR" || true
-    if [[ -n "$BUILD_PYTHONLESS" ]]; then
-        cp /$LIBTORCH_HOUSE_DIR/libtorch*.zip "$PYTORCH_FINAL_PACKAGE_DIR"
-    else
-        cp /$WHEELHOUSE_DIR/torch*.whl "$PYTORCH_FINAL_PACKAGE_DIR"
-    fi
-fi
-
-# remove stuff before testing
-rm -rf /opt/rh
-if ls /usr/local/cuda* >/dev/null 2>&1; then
-    rm -rf /usr/local/cuda*
-fi
-
-
-# Test that all the wheels work
-if [[ -z "$BUILD_PYTHONLESS" ]]; then
-  export OMP_NUM_THREADS=4 # on NUMA machines this takes too long
-  pushd $PYTORCH_ROOT/test
-
-  # Install the wheel for this Python version
-  if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-    pip uninstall -y "$TORCH_NO_PYTHON_PACKAGE_NAME" || true
-  fi
-
-  pip uninstall -y "$TORCH_PACKAGE_NAME"
-
-  if [[ "$USE_SPLIT_BUILD" == "true" ]]; then
-    pip install "$TORCH_NO_PYTHON_PACKAGE_NAME" --no-index -f /$WHEELHOUSE_DIR --no-dependencies -v
-  fi
-
-  pip install "$TORCH_PACKAGE_NAME" --no-index -f /$WHEELHOUSE_DIR --no-dependencies -v
-
-  # Print info on the libraries installed in this wheel
-  # Rather than adjust find command to skip non-library files with an embedded *.so* in their name,
-  # since this is only for reporting purposes, we add the || true to the ldd command.
-  installed_libraries=($(find "$pydir/lib/python${py_majmin}/site-packages/torch/" -name '*.so*'))
-  echo "The wheel installed all of the libraries: ${installed_libraries[@]}"
-  for installed_lib in "${installed_libraries[@]}"; do
-      ldd "$installed_lib" || true
-  done
-
-  # Run the tests
-  echo "$(date) :: Running tests"
-  pushd "$PYTORCH_ROOT"
-
-  #TODO: run_tests.sh and check_binary.sh should be moved to pytorch/pytorch project
-  LD_LIBRARY_PATH=/usr/local/nvidia/lib64 \
-          "/builder/run_tests.sh" manywheel "${py_majmin}" "$DESIRED_CUDA"
-  popd
-  echo "$(date) :: Finished tests"
-fi

.ci/manywheel/build_cpu.sh

@@ -1,99 +0,0 @@
-#!/usr/bin/env bash
-
-set -ex
-
-GPU_ARCH_TYPE=${GPU_ARCH_TYPE:-cpu}
-
-export TH_BINARY_BUILD=1
-export USE_CUDA=0
-
-# Keep an array of cmake variables to add to
-if [[ -z "$CMAKE_ARGS" ]]; then
-    # These are passed to tools/build_pytorch_libs.sh::build()
-    CMAKE_ARGS=()
-fi
-if [[ -z "$EXTRA_CAFFE2_CMAKE_FLAGS" ]]; then
-    # These are passed to tools/build_pytorch_libs.sh::build_caffe2()
-    EXTRA_CAFFE2_CMAKE_FLAGS=()
-fi
-
-DIR_SUFFIX=cpu
-if [[ "$GPU_ARCH_TYPE" == "xpu" ]]; then
-    DIR_SUFFIX=xpu
-    # Refer https://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpu/2-5.html
-    source /opt/intel/oneapi/pytorch-gpu-dev-0.5/oneapi-vars.sh
-    source /opt/intel/oneapi/pti/latest/env/vars.sh
-    export USE_STATIC_MKL=1
-fi
-
-WHEELHOUSE_DIR="wheelhouse$DIR_SUFFIX"
-LIBTORCH_HOUSE_DIR="libtorch_house$DIR_SUFFIX"
-if [[ -z "$PYTORCH_FINAL_PACKAGE_DIR" ]]; then
-    if [[ -z "$BUILD_PYTHONLESS" ]]; then
-        PYTORCH_FINAL_PACKAGE_DIR="/remote/wheelhouse$DIR_SUFFIX"
-    else
-        PYTORCH_FINAL_PACKAGE_DIR="/remote/libtorch_house$DIR_SUFFIX"
-    fi
-fi
-mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR" || true
-
-OS_NAME=$(awk -F= '/^NAME/{print $2}' /etc/os-release)
-if [[ "$OS_NAME" == *"CentOS Linux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-elif [[ "$OS_NAME" == *"Red Hat Enterprise Linux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-elif [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
-    if [[ "$(uname -m)" == "s390x" ]]; then
-        LIBGOMP_PATH="/usr/lib/s390x-linux-gnu/libgomp.so.1"
-    else
-        LIBGOMP_PATH="/usr/lib/x86_64-linux-gnu/libgomp.so.1"
-    fi
-fi
-
-DEPS_LIST=(
-    "$LIBGOMP_PATH"
-)
-
-DEPS_SONAME=(
-    "libgomp.so.1"
-)
-
-if [[ "$GPU_ARCH_TYPE" == "xpu" ]]; then
-    echo "Bundling with xpu support package libs."
-    DEPS_LIST+=(
-        "/opt/intel/oneapi/compiler/latest/lib/libsycl-preview.so.7"
-        "/opt/intel/oneapi/compiler/latest/lib/libOpenCL.so.1"
-        "/opt/intel/oneapi/compiler/latest/lib/libxptifw.so"
-        "/opt/intel/oneapi/compiler/latest/lib/libsvml.so"
-        "/opt/intel/oneapi/compiler/latest/lib/libirng.so"
-        "/opt/intel/oneapi/compiler/latest/lib/libimf.so"
-        "/opt/intel/oneapi/compiler/latest/lib/libintlc.so.5"
-        "/opt/intel/oneapi/compiler/latest/lib/libpi_level_zero.so"
-        "/opt/intel/oneapi/pti/latest/lib/libpti_view.so.0.9"
-        "/opt/intel/oneapi/pti/latest/lib/libpti.so.0.9"
-    )
-    DEPS_SONAME+=(
-        "libsycl-preview.so.7"
-        "libOpenCL.so.1"
-        "libxptifw.so"
-        "libsvml.so"
-        "libirng.so"
-        "libimf.so"
-        "libintlc.so.5"
-        "libpi_level_zero.so"
-        "libpti_view.so.0.9"
-        "libpti.so.0.9"
-    )
-fi
-
-rm -rf /usr/local/cuda*
-
-SOURCE_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null && pwd )"
-if [[ -z "$BUILD_PYTHONLESS" ]]; then
-    BUILD_SCRIPT=build_common.sh
-else
-    BUILD_SCRIPT=build_libtorch.sh
-fi
-source ${SOURCE_DIR}/${BUILD_SCRIPT}

.ci/manywheel/build_cuda.sh

@@ -1,290 +0,0 @@
-#!/usr/bin/env bash
-
-set -ex
-
-SCRIPTPATH="$( cd "$(dirname "$0")" ; pwd -P ))"
-
-export TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
-export NCCL_ROOT_DIR=/usr/local/cuda
-export TH_BINARY_BUILD=1
-export USE_STATIC_CUDNN=1
-export USE_STATIC_NCCL=1
-export ATEN_STATIC_CUDA=1
-export USE_CUDA_STATIC_LINK=1
-export INSTALL_TEST=0 # dont install test binaries into site-packages
-export USE_CUPTI_SO=0
-export USE_CUSPARSELT=${USE_CUSPARSELT:-1} # Enable if not disabled by libtorch build
-
-# Keep an array of cmake variables to add to
-if [[ -z "$CMAKE_ARGS" ]]; then
-    # These are passed to tools/build_pytorch_libs.sh::build()
-    CMAKE_ARGS=()
-fi
-if [[ -z "$EXTRA_CAFFE2_CMAKE_FLAGS" ]]; then
-    # These are passed to tools/build_pytorch_libs.sh::build_caffe2()
-    EXTRA_CAFFE2_CMAKE_FLAGS=()
-fi
-
-# Determine CUDA version and architectures to build for
-#
-# NOTE: We should first check `DESIRED_CUDA` when determining `CUDA_VERSION`,
-# because in some cases a single Docker image can have multiple CUDA versions
-# on it, and `nvcc --version` might not show the CUDA version we want.
-if [[ -n "$DESIRED_CUDA" ]]; then
-    # If the DESIRED_CUDA already matches the format that we expect
-    if [[ ${DESIRED_CUDA} =~ ^[0-9]+\.[0-9]+$ ]]; then
-        CUDA_VERSION=${DESIRED_CUDA}
-    else
-        # cu90, cu92, cu100, cu101
-        if [[ ${#DESIRED_CUDA} -eq 4 ]]; then
-            CUDA_VERSION="${DESIRED_CUDA:2:1}.${DESIRED_CUDA:3:1}"
-        elif [[ ${#DESIRED_CUDA} -eq 5 ]]; then
-            CUDA_VERSION="${DESIRED_CUDA:2:2}.${DESIRED_CUDA:4:1}"
-        fi
-    fi
-    echo "Using CUDA $CUDA_VERSION as determined by DESIRED_CUDA"
-
-    # There really has to be a better way to do this - eli
-    # Possibly limiting builds to specific cuda versions be delimiting images would be a choice
-    if [[ "$OS_NAME" == *"Ubuntu"* ]]; then
-        echo "Switching to CUDA version ${DESIRED_CUDA}"
-        /builder/conda/switch_cuda_version.sh "${DESIRED_CUDA}"
-    fi
-else
-    CUDA_VERSION=$(nvcc --version|grep release|cut -f5 -d" "|cut -f1 -d",")
-    echo "CUDA $CUDA_VERSION Detected"
-fi
-
-cuda_version_nodot=$(echo $CUDA_VERSION | tr -d '.')
-
-TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6"
-case ${CUDA_VERSION} in
-    12.4)
-        if [[ "$GPU_ARCH_TYPE" = "cuda-aarch64" ]]; then
-            TORCH_CUDA_ARCH_LIST="9.0"
-        else
-            TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0+PTX"
-        fi
-        EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
-        ;;
-    12.1)
-        TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0"
-        EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
-        ;;
-    11.8)
-        TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};3.7;9.0"
-        EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
-        ;;
-    11.[67])
-        TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};3.7"
-        EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
-        ;;
-    *)
-        echo "unknown cuda version $CUDA_VERSION"
-        exit 1
-        ;;
-esac
-
-export TORCH_CUDA_ARCH_LIST=${TORCH_CUDA_ARCH_LIST}
-echo "${TORCH_CUDA_ARCH_LIST}"
-
-# Package directories
-WHEELHOUSE_DIR="wheelhouse$cuda_version_nodot"
-LIBTORCH_HOUSE_DIR="libtorch_house$cuda_version_nodot"
-if [[ -z "$PYTORCH_FINAL_PACKAGE_DIR" ]]; then
-    if [[ -z "$BUILD_PYTHONLESS" ]]; then
-        PYTORCH_FINAL_PACKAGE_DIR="/remote/wheelhouse$cuda_version_nodot"
-    else
-        PYTORCH_FINAL_PACKAGE_DIR="/remote/libtorch_house$cuda_version_nodot"
-    fi
-fi
-mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR" || true
-
-OS_NAME=$(awk -F= '/^NAME/{print $2}' /etc/os-release)
-if [[ "$OS_NAME" == *"CentOS Linux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-elif [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-elif [[ "$OS_NAME" == *"Red Hat Enterprise Linux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
-    LIBGOMP_PATH="/usr/lib/x86_64-linux-gnu/libgomp.so.1"
-fi
-
-DEPS_LIST=(
-    "$LIBGOMP_PATH"
-)
-DEPS_SONAME=(
-    "libgomp.so.1"
-)
-
-if [[ $USE_CUSPARSELT == "1" ]]; then
-        DEPS_SONAME+=(
-            "libcusparseLt.so.0"
-        )
-        DEPS_LIST+=(
-            "/usr/local/cuda/lib64/libcusparseLt.so.0"
-        )
-fi
-
-if [[ $CUDA_VERSION == "12.1" || $CUDA_VERSION == "12.4" ]]; then
-    export USE_STATIC_CUDNN=0
-    # Try parallelizing nvcc as well
-    export TORCH_NVCC_FLAGS="-Xfatbin -compress-all --threads 2"
-
-    if [[ -z "$PYTORCH_EXTRA_INSTALL_REQUIREMENTS" ]]; then
-        echo "Bundling with cudnn and cublas."
-        DEPS_LIST+=(
-            "/usr/local/cuda/lib64/libcudnn_adv.so.9"
-            "/usr/local/cuda/lib64/libcudnn_cnn.so.9"
-            "/usr/local/cuda/lib64/libcudnn_graph.so.9"
-            "/usr/local/cuda/lib64/libcudnn_ops.so.9"
-            "/usr/local/cuda/lib64/libcudnn_engines_runtime_compiled.so.9"
-            "/usr/local/cuda/lib64/libcudnn_engines_precompiled.so.9"
-            "/usr/local/cuda/lib64/libcudnn_heuristic.so.9"
-            "/usr/local/cuda/lib64/libcudnn.so.9"
-            "/usr/local/cuda/lib64/libcublas.so.12"
-            "/usr/local/cuda/lib64/libcublasLt.so.12"
-            "/usr/local/cuda/lib64/libcudart.so.12"
-            "/usr/local/cuda/lib64/libnvToolsExt.so.1"
-            "/usr/local/cuda/lib64/libnvrtc.so.12"
-            "/usr/local/cuda/lib64/libnvrtc-builtins.so"
-        )
-        DEPS_SONAME+=(
-            "libcudnn_adv.so.9"
-            "libcudnn_cnn.so.9"
-            "libcudnn_graph.so.9"
-            "libcudnn_ops.so.9"
-            "libcudnn_engines_runtime_compiled.so.9"
-            "libcudnn_engines_precompiled.so.9"
-            "libcudnn_heuristic.so.9"
-            "libcudnn.so.9"
-            "libcublas.so.12"
-            "libcublasLt.so.12"
-            "libcudart.so.12"
-            "libnvToolsExt.so.1"
-            "libnvrtc.so.12"
-            "libnvrtc-builtins.so"
-        )
-    else
-        echo "Using nvidia libs from pypi."
-        CUDA_RPATHS=(
-            '$ORIGIN/../../nvidia/cublas/lib'
-            '$ORIGIN/../../nvidia/cuda_cupti/lib'
-            '$ORIGIN/../../nvidia/cuda_nvrtc/lib'
-            '$ORIGIN/../../nvidia/cuda_runtime/lib'
-            '$ORIGIN/../../nvidia/cudnn/lib'
-            '$ORIGIN/../../nvidia/cufft/lib'
-            '$ORIGIN/../../nvidia/curand/lib'
-            '$ORIGIN/../../nvidia/cusolver/lib'
-            '$ORIGIN/../../nvidia/cusparse/lib'
-            '$ORIGIN/../../nvidia/nccl/lib'
-            '$ORIGIN/../../nvidia/nvtx/lib'
-        )
-        CUDA_RPATHS=$(IFS=: ; echo "${CUDA_RPATHS[*]}")
-        export C_SO_RPATH=$CUDA_RPATHS':$ORIGIN:$ORIGIN/lib'
-        export LIB_SO_RPATH=$CUDA_RPATHS':$ORIGIN'
-        export FORCE_RPATH="--force-rpath"
-        export USE_STATIC_NCCL=0
-        export USE_SYSTEM_NCCL=1
-        export ATEN_STATIC_CUDA=0
-        export USE_CUDA_STATIC_LINK=0
-        export USE_CUPTI_SO=1
-        export NCCL_INCLUDE_DIR="/usr/local/cuda/include/"
-        export NCCL_LIB_DIR="/usr/local/cuda/lib64/"
-    fi
-elif [[ $CUDA_VERSION == "11.8" ]]; then
-    export USE_STATIC_CUDNN=0
-    # Try parallelizing nvcc as well
-    export TORCH_NVCC_FLAGS="-Xfatbin -compress-all --threads 2"
-    # Bundle ptxas into the wheel, see https://github.com/pytorch/pytorch/pull/119750
-    export BUILD_BUNDLE_PTXAS=1
-
-    if [[ -z "$PYTORCH_EXTRA_INSTALL_REQUIREMENTS" ]]; then
-        echo "Bundling with cudnn and cublas."
-        DEPS_LIST+=(
-            "/usr/local/cuda/lib64/libcudnn_adv.so.9"
-            "/usr/local/cuda/lib64/libcudnn_cnn.so.9"
-            "/usr/local/cuda/lib64/libcudnn_graph.so.9"
-            "/usr/local/cuda/lib64/libcudnn_ops.so.9"
-            "/usr/local/cuda/lib64/libcudnn_engines_runtime_compiled.so.9"
-            "/usr/local/cuda/lib64/libcudnn_engines_precompiled.so.9"
-            "/usr/local/cuda/lib64/libcudnn_heuristic.so.9"
-            "/usr/local/cuda/lib64/libcudnn.so.9"
-            "/usr/local/cuda/lib64/libcublas.so.11"
-            "/usr/local/cuda/lib64/libcublasLt.so.11"
-            "/usr/local/cuda/lib64/libcudart.so.11.0"
-            "/usr/local/cuda/lib64/libnvToolsExt.so.1"
-            "/usr/local/cuda/lib64/libnvrtc.so.11.2"    # this is not a mistake, it links to more specific cuda version
-            "/usr/local/cuda/lib64/libnvrtc-builtins.so.11.8"
-        )
-        DEPS_SONAME+=(
-            "libcudnn_adv.so.9"
-            "libcudnn_cnn.so.9"
-            "libcudnn_graph.so.9"
-            "libcudnn_ops.so.9"
-            "libcudnn_engines_runtime_compiled.so.9"
-            "libcudnn_engines_precompiled.so.9"
-            "libcudnn_heuristic.so.9"
-            "libcudnn.so.9"
-            "libcublas.so.11"
-            "libcublasLt.so.11"
-            "libcudart.so.11.0"
-            "libnvToolsExt.so.1"
-            "libnvrtc.so.11.2"
-            "libnvrtc-builtins.so.11.8"
-        )
-    else
-        echo "Using nvidia libs from pypi."
-        CUDA_RPATHS=(
-            '$ORIGIN/../../nvidia/cublas/lib'
-            '$ORIGIN/../../nvidia/cuda_cupti/lib'
-            '$ORIGIN/../../nvidia/cuda_nvrtc/lib'
-            '$ORIGIN/../../nvidia/cuda_runtime/lib'
-            '$ORIGIN/../../nvidia/cudnn/lib'
-            '$ORIGIN/../../nvidia/cufft/lib'
-            '$ORIGIN/../../nvidia/curand/lib'
-            '$ORIGIN/../../nvidia/cusolver/lib'
-            '$ORIGIN/../../nvidia/cusparse/lib'
-            '$ORIGIN/../../nvidia/nccl/lib'
-            '$ORIGIN/../../nvidia/nvtx/lib'
-        )
-        CUDA_RPATHS=$(IFS=: ; echo "${CUDA_RPATHS[*]}")
-        export C_SO_RPATH=$CUDA_RPATHS':$ORIGIN:$ORIGIN/lib'
-        export LIB_SO_RPATH=$CUDA_RPATHS':$ORIGIN'
-        export FORCE_RPATH="--force-rpath"
-        export USE_STATIC_NCCL=0
-        export USE_SYSTEM_NCCL=1
-        export ATEN_STATIC_CUDA=0
-        export USE_CUDA_STATIC_LINK=0
-        export USE_CUPTI_SO=1
-        export NCCL_INCLUDE_DIR="/usr/local/cuda/include/"
-        export NCCL_LIB_DIR="/usr/local/cuda/lib64/"
-    fi
-else
-    echo "Unknown cuda version $CUDA_VERSION"
-    exit 1
-fi
-
-# builder/test.sh requires DESIRED_CUDA to know what tests to exclude
-export DESIRED_CUDA="$cuda_version_nodot"
-
-# Switch `/usr/local/cuda` to the desired CUDA version
-rm -rf /usr/local/cuda || true
-ln -s "/usr/local/cuda-${CUDA_VERSION}" /usr/local/cuda
-
-# Switch `/usr/local/magma` to the desired CUDA version
-rm -rf /usr/local/magma || true
-ln -s /usr/local/cuda-${CUDA_VERSION}/magma /usr/local/magma
-
-export CUDA_VERSION=$(ls /usr/local/cuda/lib64/libcudart.so.*|sort|tac | head -1 | rev | cut -d"." -f -3 | rev) # 10.0.130
-export CUDA_VERSION_SHORT=$(ls /usr/local/cuda/lib64/libcudart.so.*|sort|tac | head -1 | rev | cut -d"." -f -3 | rev | cut -f1,2 -d".") # 10.0
-export CUDNN_VERSION=$(ls /usr/local/cuda/lib64/libcudnn.so.*|sort|tac | head -1 | rev | cut -d"." -f -3 | rev)
-
-SCRIPTPATH="$( cd "$(dirname "$0")" ; pwd -P )"
-if [[ -z "$BUILD_PYTHONLESS" ]]; then
-    BUILD_SCRIPT=build_common.sh
-else
-    BUILD_SCRIPT=build_libtorch.sh
-fi
-source $SCRIPTPATH/${BUILD_SCRIPT}

.ci/manywheel/build_libtorch.sh

@@ -1,353 +0,0 @@
-#!/usr/bin/env bash
-# meant to be called only from the neighboring build.sh and build_cpu.sh scripts
-
-set -e pipefail
-SOURCE_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null && pwd )"
-
-# Require only one python installation
-if [[ -z "$DESIRED_PYTHON" ]]; then
-    echo "Need to set DESIRED_PYTHON env variable"
-    exit 1
-fi
-if [[ -n "$BUILD_PYTHONLESS" && -z "$LIBTORCH_VARIANT" ]]; then
-    echo "BUILD_PYTHONLESS is set, so need LIBTORCH_VARIANT to also be set"
-    echo "LIBTORCH_VARIANT should be one of shared-with-deps shared-without-deps static-with-deps static-without-deps"
-    exit 1
-fi
-
-# Function to retry functions that sometimes timeout or have flaky failures
-retry () {
-    $*  || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
-}
-
-# TODO move this into the Docker images
-OS_NAME=`awk -F= '/^NAME/{print $2}' /etc/os-release`
-if [[ "$OS_NAME" == *"CentOS Linux"* ]]; then
-    retry yum install -q -y zip openssl
-elif [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
-    retry yum install -q -y zip openssl
-elif [[ "$OS_NAME" == *"Red Hat Enterprise Linux"* ]]; then
-    retry dnf install -q -y zip openssl
-elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
-    # TODO: Remove this once nvidia package repos are back online
-    # Comment out nvidia repositories to prevent them from getting apt-get updated, see https://github.com/pytorch/pytorch/issues/74968
-    # shellcheck disable=SC2046
-    sed -i 's/.*nvidia.*/# &/' $(find /etc/apt/ -type f -name "*.list")
-    retry apt-get update
-    retry apt-get -y install zip openssl
-fi
-
-# Version: setup.py uses $PYTORCH_BUILD_VERSION.post$PYTORCH_BUILD_NUMBER if
-# PYTORCH_BUILD_NUMBER > 1
-build_version="$PYTORCH_BUILD_VERSION"
-build_number="$PYTORCH_BUILD_NUMBER"
-if [[ -n "$OVERRIDE_PACKAGE_VERSION" ]]; then
-    # This will be the *exact* version, since build_number<1
-    build_version="$OVERRIDE_PACKAGE_VERSION"
-    build_number=0
-fi
-if [[ -z "$build_version" ]]; then
-    build_version=1.0.0
-fi
-if [[ -z "$build_number" ]]; then
-    build_number=1
-fi
-export PYTORCH_BUILD_VERSION=$build_version
-export PYTORCH_BUILD_NUMBER=$build_number
-
-export CMAKE_LIBRARY_PATH="/opt/intel/lib:/lib:$CMAKE_LIBRARY_PATH"
-export CMAKE_INCLUDE_PATH="/opt/intel/include:$CMAKE_INCLUDE_PATH"
-
-# set OPENSSL_ROOT_DIR=/opt/openssl if it exists
-if [[ -e /opt/openssl ]]; then
-    export OPENSSL_ROOT_DIR=/opt/openssl
-    export CMAKE_INCLUDE_PATH="/opt/openssl/include":$CMAKE_INCLUDE_PATH
-fi
-
-# If given a python version like 3.6m or 2.7mu, convert this to the format we
-# expect. The binary CI jobs pass in python versions like this; they also only
-# ever pass one python version, so we assume that DESIRED_PYTHON is not a list
-# in this case
-if [[ -n "$DESIRED_PYTHON" && "$DESIRED_PYTHON" != cp* ]]; then
-    python_nodot="$(echo $DESIRED_PYTHON | tr -d m.u)"
-    DESIRED_PYTHON="cp${python_nodot}-cp${python_nodot}"
-fi
-pydir="/opt/python/$DESIRED_PYTHON"
-export PATH="$pydir/bin:$PATH"
-
-export PATCHELF_BIN=/usr/local/bin/patchelf
-patchelf_version=`$PATCHELF_BIN --version`
-echo "patchelf version: " $patchelf_version
-if [[ "$patchelf_version" == "patchelf 0.9" ]]; then
-    echo "Your patchelf version is too old. Please use version >= 0.10."
-    exit 1
-fi
-
-########################################################
-# Compile wheels as well as libtorch
-#######################################################
-if [[ -z "$PYTORCH_ROOT" ]]; then
-    echo "Need to set PYTORCH_ROOT env variable"
-    exit 1
-fi
-pushd "$PYTORCH_ROOT"
-python setup.py clean
-retry pip install -qr requirements.txt
-retry pip install -q numpy==2.0.1
-
-if [[ "$DESIRED_DEVTOOLSET" == *"cxx11-abi"* ]]; then
-    export _GLIBCXX_USE_CXX11_ABI=1
-else
-    export _GLIBCXX_USE_CXX11_ABI=0
-fi
-
-if [[ "$DESIRED_CUDA" == *"rocm"* ]]; then
-    echo "Calling build_amd.py at $(date)"
-    python tools/amd_build/build_amd.py
-    # TODO remove this work-around once pytorch sources are updated
-    export ROCclr_DIR=/opt/rocm/rocclr/lib/cmake/rocclr
-fi
-
-echo "Calling setup.py install at $(date)"
-
-if [[ $LIBTORCH_VARIANT = *"static"* ]]; then
-    STATIC_CMAKE_FLAG="-DTORCH_STATIC=1"
-fi
-
-(
-    set -x
-
-    mkdir -p build
-
-    time CMAKE_ARGS=${CMAKE_ARGS[@]} \
-        EXTRA_CAFFE2_CMAKE_FLAGS="${EXTRA_CAFFE2_CMAKE_FLAGS[@]} $STATIC_CMAKE_FLAG" \
-        # TODO: Remove this flag once https://github.com/pytorch/pytorch/issues/55952 is closed
-        CFLAGS='-Wno-deprecated-declarations' \
-        BUILD_LIBTORCH_CPU_WITH_DEBUG=1 \
-        python setup.py install
-
-    mkdir -p libtorch/{lib,bin,include,share}
-
-    # Make debug folder separate so it doesn't get zipped up with the rest of
-    # libtorch
-    mkdir debug
-
-    # Copy over all lib files
-    cp -rv build/lib/*                libtorch/lib/
-    cp -rv build/lib*/torch/lib/*     libtorch/lib/
-
-    # Copy over all include files
-    cp -rv build/include/*            libtorch/include/
-    cp -rv build/lib*/torch/include/* libtorch/include/
-
-    # Copy over all of the cmake files
-    cp -rv build/lib*/torch/share/*   libtorch/share/
-
-    # Split libtorch into debug / release version
-    cp libtorch/lib/libtorch_cpu.so libtorch/lib/libtorch_cpu.so.dbg
-
-    # Keep debug symbols on debug lib
-    strip --only-keep-debug libtorch/lib/libtorch_cpu.so.dbg
-
-    # Remove debug info from release lib
-    strip --strip-debug libtorch/lib/libtorch_cpu.so
-
-    # Add a debug link to the release lib to the debug lib (debuggers will then
-    # search for symbols in a file called libtorch_cpu.so.dbg in some
-    # predetermined locations) and embed a CRC32 of the debug library into the .so
-    cd libtorch/lib
-
-    objcopy libtorch_cpu.so --add-gnu-debuglink=libtorch_cpu.so.dbg
-    cd ../..
-
-    # Move the debug symbols to its own directory so it doesn't get processed /
-    # zipped with all the other libraries
-    mv libtorch/lib/libtorch_cpu.so.dbg debug/libtorch_cpu.so.dbg
-
-    echo "${PYTORCH_BUILD_VERSION}" > libtorch/build-version
-    echo "$(pushd $PYTORCH_ROOT && git rev-parse HEAD)" > libtorch/build-hash
-
-)
-
-if [[ "$DESIRED_DEVTOOLSET" == *"cxx11-abi"* ]]; then
-    LIBTORCH_ABI="cxx11-abi-"
-else
-    LIBTORCH_ABI=
-fi
-
-(
-    set -x
-
-    mkdir -p /tmp/$LIBTORCH_HOUSE_DIR
-
-    # objcopy installs a CRC32 into libtorch_cpu above so, so add that to the name here
-    CRC32=$(objcopy --dump-section .gnu_debuglink=>(tail -c4 | od -t x4 -An | xargs echo) libtorch/lib/libtorch_cpu.so)
-
-    # Zip debug symbols
-    zip /tmp/$LIBTORCH_HOUSE_DIR/debug-libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-$PYTORCH_BUILD_VERSION-$CRC32.zip debug/libtorch_cpu.so.dbg
-
-    # Zip and copy libtorch
-    zip -rq /tmp/$LIBTORCH_HOUSE_DIR/libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-$PYTORCH_BUILD_VERSION.zip libtorch
-    cp /tmp/$LIBTORCH_HOUSE_DIR/libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-$PYTORCH_BUILD_VERSION.zip \
-       /tmp/$LIBTORCH_HOUSE_DIR/libtorch-$LIBTORCH_ABI$LIBTORCH_VARIANT-latest.zip
-)
-
-
-popd
-
-#######################################################################
-# ADD DEPENDENCIES INTO THE WHEEL
-#
-# auditwheel repair doesn't work correctly and is buggy
-# so manually do the work of copying dependency libs and patchelfing
-# and fixing RECORDS entries correctly
-######################################################################
-
-fname_with_sha256() {
-    HASH=$(sha256sum $1 | cut -c1-8)
-    DIRNAME=$(dirname $1)
-    BASENAME=$(basename $1)
-    if [[ $BASENAME == "libnvrtc-builtins.so" || $BASENAME == "libcudnn"* ]]; then
-        echo $1
-    else
-        INITNAME=$(echo $BASENAME | cut -f1 -d".")
-        ENDNAME=$(echo $BASENAME | cut -f 2- -d".")
-        echo "$DIRNAME/$INITNAME-$HASH.$ENDNAME"
-    fi
-}
-
-fname_without_so_number() {
-    LINKNAME=$(echo $1 | sed -e 's/\.so.*/.so/g')
-    echo "$LINKNAME"
-}
-
-make_wheel_record() {
-    FPATH=$1
-    if echo $FPATH | grep RECORD >/dev/null 2>&1; then
-        # if the RECORD file, then
-        echo "$FPATH,,"
-    else
-        HASH=$(openssl dgst -sha256 -binary $FPATH | openssl base64 | sed -e 's/+/-/g' | sed -e 's/\//_/g' | sed -e 's/=//g')
-        FSIZE=$(ls -nl $FPATH | awk '{print $5}')
-        echo "$FPATH,sha256=$HASH,$FSIZE"
-    fi
-}
-
-echo 'Built this package:'
-(
-    set -x
-    mkdir -p /$LIBTORCH_HOUSE_DIR
-    mv /tmp/$LIBTORCH_HOUSE_DIR/*.zip /$LIBTORCH_HOUSE_DIR
-    rm -rf /tmp/$LIBTORCH_HOUSE_DIR
-)
-TMP_DIR=$(mktemp -d)
-trap "rm -rf ${TMP_DIR}" EXIT
-pushd "${TMP_DIR}"
-
-for pkg in /$LIBTORCH_HOUSE_DIR/libtorch*.zip; do
-
-    # if the glob didn't match anything
-    if [[ ! -e $pkg ]]; then
-        continue
-    fi
-
-    rm -rf tmp
-    mkdir -p tmp
-    cd tmp
-    cp $pkg .
-
-    unzip -q $(basename $pkg)
-    rm -f $(basename $pkg)
-
-    PREFIX=libtorch
-
-    if [[ $pkg != *"without-deps"* ]]; then
-        # copy over needed dependent .so files over and tag them with their hash
-        patched=()
-        for filepath in "${DEPS_LIST[@]}"; do
-            filename=$(basename $filepath)
-            destpath=$PREFIX/lib/$filename
-            if [[ "$filepath" != "$destpath" ]]; then
-                cp $filepath $destpath
-            fi
-
-            if [[ "$DESIRED_CUDA" == *"rocm"* ]]; then
-                patchedpath=$(fname_without_so_number $destpath)
-            else
-                patchedpath=$(fname_with_sha256 $destpath)
-            fi
-            patchedname=$(basename $patchedpath)
-            if [[ "$destpath" != "$patchedpath" ]]; then
-                mv $destpath $patchedpath
-            fi
-            patched+=("$patchedname")
-            echo "Copied $filepath to $patchedpath"
-        done
-
-        echo "patching to fix the so names to the hashed names"
-        for ((i=0;i<${#DEPS_LIST[@]};++i)); do
-            find $PREFIX -name '*.so*' | while read sofile; do
-                origname=${DEPS_SONAME[i]}
-                patchedname=${patched[i]}
-                if [[ "$origname" != "$patchedname" ]] || [[ "$DESIRED_CUDA" == *"rocm"* ]]; 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 $sofile
-                    fi
-                fi
-            done
-        done
-
-        # copy over needed auxiliary files
-        for ((i=0;i<${#DEPS_AUX_SRCLIST[@]};++i)); do
-            srcpath=${DEPS_AUX_SRCLIST[i]}
-            dstpath=$PREFIX/${DEPS_AUX_DSTLIST[i]}
-            mkdir -p $(dirname $dstpath)
-            cp $srcpath $dstpath
-        done
-    fi
-
-    # set RPATH of _C.so and similar to $ORIGIN, $ORIGIN/lib
-    find $PREFIX -maxdepth 1 -type f -name "*.so*" | while read sofile; do
-        echo "Setting rpath of $sofile to " '$ORIGIN:$ORIGIN/lib'
-        $PATCHELF_BIN --set-rpath '$ORIGIN:$ORIGIN/lib' $sofile
-        $PATCHELF_BIN --print-rpath $sofile
-    done
-
-    # set RPATH of lib/ files to $ORIGIN
-    find $PREFIX/lib -maxdepth 1 -type f -name "*.so*" | while read sofile; do
-        echo "Setting rpath of $sofile to " '$ORIGIN'
-        $PATCHELF_BIN --set-rpath '$ORIGIN' $sofile
-        $PATCHELF_BIN --print-rpath $sofile
-    done
-
-    # regenerate the RECORD file with new hashes
-    record_file=`echo $(basename $pkg) | sed -e 's/-cp.*$/.dist-info\/RECORD/g'`
-    if [[ -e $record_file ]]; then
-        echo "Generating new record file $record_file"
-        rm -f $record_file
-        # generate records for folders in wheel
-        find * -type f | while read fname; do
-            echo $(make_wheel_record $fname) >>$record_file
-        done
-    fi
-
-    # zip up the wheel back
-    zip -rq $(basename $pkg) $PREFIX*
-
-    # replace original wheel
-    rm -f $pkg
-    mv $(basename $pkg) $pkg
-    cd ..
-    rm -rf tmp
-done
-
-# Copy wheels to host machine for persistence before testing
-if [[ -n "$PYTORCH_FINAL_PACKAGE_DIR" ]]; then
-    cp /$LIBTORCH_HOUSE_DIR/libtorch*.zip "$PYTORCH_FINAL_PACKAGE_DIR"
-    cp /$LIBTORCH_HOUSE_DIR/debug-libtorch*.zip "$PYTORCH_FINAL_PACKAGE_DIR"
-fi

.ci/manywheel/build_rocm.sh

@@ -1,263 +0,0 @@
-#!/usr/bin/env bash
-
-set -ex
-
-export ROCM_HOME=/opt/rocm
-export MAGMA_HOME=$ROCM_HOME/magma
-# TODO: libtorch_cpu.so is broken when building with Debug info
-export BUILD_DEBUG_INFO=0
-
-# TODO Are these all used/needed?
-export TH_BINARY_BUILD=1
-export USE_STATIC_CUDNN=1
-export USE_STATIC_NCCL=1
-export ATEN_STATIC_CUDA=1
-export USE_CUDA_STATIC_LINK=1
-export INSTALL_TEST=0 # dont install test binaries into site-packages
-# Set RPATH instead of RUNPATH when using patchelf to avoid LD_LIBRARY_PATH override
-export FORCE_RPATH="--force-rpath"
-
-# Keep an array of cmake variables to add to
-if [[ -z "$CMAKE_ARGS" ]]; then
-    # These are passed to tools/build_pytorch_libs.sh::build()
-    CMAKE_ARGS=()
-fi
-if [[ -z "$EXTRA_CAFFE2_CMAKE_FLAGS" ]]; then
-    # These are passed to tools/build_pytorch_libs.sh::build_caffe2()
-    EXTRA_CAFFE2_CMAKE_FLAGS=()
-fi
-
-# Determine ROCm version and architectures to build for
-#
-# NOTE: We should first check `DESIRED_CUDA` when determining `ROCM_VERSION`
-if [[ -n "$DESIRED_CUDA" ]]; then
-    if ! echo "${DESIRED_CUDA}"| grep "^rocm" >/dev/null 2>/dev/null; then
-        export DESIRED_CUDA="rocm${DESIRED_CUDA}"
-    fi
-    # rocm3.7, rocm3.5.1
-    ROCM_VERSION="$DESIRED_CUDA"
-    echo "Using $ROCM_VERSION as determined by DESIRED_CUDA"
-else
-    echo "Must set DESIRED_CUDA"
-    exit 1
-fi
-
-# Package directories
-WHEELHOUSE_DIR="wheelhouse$ROCM_VERSION"
-LIBTORCH_HOUSE_DIR="libtorch_house$ROCM_VERSION"
-if [[ -z "$PYTORCH_FINAL_PACKAGE_DIR" ]]; then
-    if [[ -z "$BUILD_PYTHONLESS" ]]; then
-        PYTORCH_FINAL_PACKAGE_DIR="/remote/wheelhouse$ROCM_VERSION"
-    else
-        PYTORCH_FINAL_PACKAGE_DIR="/remote/libtorch_house$ROCM_VERSION"
-    fi
-fi
-mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR" || true
-
-# To make version comparison easier, create an integer representation.
-ROCM_VERSION_CLEAN=$(echo ${ROCM_VERSION} | sed s/rocm//)
-save_IFS="$IFS"
-IFS=. ROCM_VERSION_ARRAY=(${ROCM_VERSION_CLEAN})
-IFS="$save_IFS"
-if [[ ${#ROCM_VERSION_ARRAY[@]} == 2 ]]; then
-    ROCM_VERSION_MAJOR=${ROCM_VERSION_ARRAY[0]}
-    ROCM_VERSION_MINOR=${ROCM_VERSION_ARRAY[1]}
-    ROCM_VERSION_PATCH=0
-elif [[ ${#ROCM_VERSION_ARRAY[@]} == 3 ]]; then
-    ROCM_VERSION_MAJOR=${ROCM_VERSION_ARRAY[0]}
-    ROCM_VERSION_MINOR=${ROCM_VERSION_ARRAY[1]}
-    ROCM_VERSION_PATCH=${ROCM_VERSION_ARRAY[2]}
-else
-    echo "Unhandled ROCM_VERSION ${ROCM_VERSION}"
-    exit 1
-fi
-ROCM_INT=$(($ROCM_VERSION_MAJOR * 10000 + $ROCM_VERSION_MINOR * 100 + $ROCM_VERSION_PATCH))
-
-# Required ROCm libraries
-ROCM_SO_FILES=(
-    "libMIOpen.so"
-    "libamdhip64.so"
-    "libhipblas.so"
-    "libhipfft.so"
-    "libhiprand.so"
-    "libhipsolver.so"
-    "libhipsparse.so"
-    "libhsa-runtime64.so"
-    "libamd_comgr.so"
-    "libmagma.so"
-    "librccl.so"
-    "librocblas.so"
-    "librocfft.so"
-    "librocm_smi64.so"
-    "librocrand.so"
-    "librocsolver.so"
-    "librocsparse.so"
-    "libroctracer64.so"
-    "libroctx64.so"
-    "libhipblaslt.so"
-    "libhiprtc.so"
-)
-
-if [[ $ROCM_INT -ge 60100 ]]; then
-    ROCM_SO_FILES+=("librocprofiler-register.so")
-fi
-
-if [[ $ROCM_INT -ge 60200 ]]; then
-    ROCM_SO_FILES+=("librocm-core.so")
-fi
-
-OS_NAME=`awk -F= '/^NAME/{print $2}' /etc/os-release`
-if [[ "$OS_NAME" == *"CentOS Linux"* ]]; then
-    LIBGOMP_PATH="/usr/lib64/libgomp.so.1"
-    LIBNUMA_PATH="/usr/lib64/libnuma.so.1"
-    LIBELF_PATH="/usr/lib64/libelf.so.1"
-    LIBTINFO_PATH="/usr/lib64/libtinfo.so.5"
-    LIBDRM_PATH="/opt/amdgpu/lib64/libdrm.so.2"
-    LIBDRM_AMDGPU_PATH="/opt/amdgpu/lib64/libdrm_amdgpu.so.1"
-    if [[ $ROCM_INT -ge 60100 ]]; then
-        # Below libs are direct dependencies of libhipsolver
-        LIBSUITESPARSE_CONFIG_PATH="/lib64/libsuitesparseconfig.so.4"
-        LIBCHOLMOD_PATH="/lib64/libcholmod.so.2"
-        # Below libs are direct dependencies of libcholmod
-        LIBAMD_PATH="/lib64/libamd.so.2"
-        LIBCAMD_PATH="/lib64/libcamd.so.2"
-        LIBCCOLAMD_PATH="/lib64/libccolamd.so.2"
-        LIBCOLAMD_PATH="/lib64/libcolamd.so.2"
-        LIBSATLAS_PATH="/lib64/atlas/libsatlas.so.3"
-        # Below libs are direct dependencies of libsatlas
-        LIBGFORTRAN_PATH="/lib64/libgfortran.so.3"
-        LIBQUADMATH_PATH="/lib64/libquadmath.so.0"
-    fi
-    MAYBE_LIB64=lib64
-elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
-    LIBGOMP_PATH="/usr/lib/x86_64-linux-gnu/libgomp.so.1"
-    LIBNUMA_PATH="/usr/lib/x86_64-linux-gnu/libnuma.so.1"
-    LIBELF_PATH="/usr/lib/x86_64-linux-gnu/libelf.so.1"
-    if [[ $ROCM_INT -ge 50300 ]]; then
-        LIBTINFO_PATH="/lib/x86_64-linux-gnu/libtinfo.so.6"
-    else
-        LIBTINFO_PATH="/lib/x86_64-linux-gnu/libtinfo.so.5"
-    fi
-    LIBDRM_PATH="/usr/lib/x86_64-linux-gnu/libdrm.so.2"
-    LIBDRM_AMDGPU_PATH="/usr/lib/x86_64-linux-gnu/libdrm_amdgpu.so.1"
-    if [[ $ROCM_INT -ge 60100 ]]; then
-        # Below libs are direct dependencies of libhipsolver
-        LIBCHOLMOD_PATH="/lib/x86_64-linux-gnu/libcholmod.so.3"
-        # Below libs are direct dependencies of libcholmod
-        LIBSUITESPARSE_CONFIG_PATH="/lib/x86_64-linux-gnu/libsuitesparseconfig.so.5"
-        LIBAMD_PATH="/lib/x86_64-linux-gnu/libamd.so.2"
-        LIBCAMD_PATH="/lib/x86_64-linux-gnu/libcamd.so.2"
-        LIBCCOLAMD_PATH="/lib/x86_64-linux-gnu/libccolamd.so.2"
-        LIBCOLAMD_PATH="/lib/x86_64-linux-gnu/libcolamd.so.2"
-        LIBMETIS_PATH="/lib/x86_64-linux-gnu/libmetis.so.5"
-        LIBLAPACK_PATH="/lib/x86_64-linux-gnu/liblapack.so.3"
-        LIBBLAS_PATH="/lib/x86_64-linux-gnu/libblas.so.3"
-        # Below libs are direct dependencies of libblas
-        LIBGFORTRAN_PATH="/lib/x86_64-linux-gnu/libgfortran.so.5"
-        LIBQUADMATH_PATH="/lib/x86_64-linux-gnu/libquadmath.so.0"
-    fi
-    MAYBE_LIB64=lib
-fi
-OS_SO_PATHS=($LIBGOMP_PATH $LIBNUMA_PATH\
-             $LIBELF_PATH $LIBTINFO_PATH\
-             $LIBDRM_PATH $LIBDRM_AMDGPU_PATH\
-             $LIBSUITESPARSE_CONFIG_PATH\
-             $LIBCHOLMOD_PATH $LIBAMD_PATH\
-             $LIBCAMD_PATH $LIBCCOLAMD_PATH\
-             $LIBCOLAMD_PATH $LIBSATLAS_PATH\
-             $LIBGFORTRAN_PATH $LIBQUADMATH_PATH\
-             $LIBMETIS_PATH $LIBLAPACK_PATH\
-             $LIBBLAS_PATH)
-OS_SO_FILES=()
-for lib in "${OS_SO_PATHS[@]}"
-do
-    file_name="${lib##*/}" # Substring removal of path to get filename
-    OS_SO_FILES[${#OS_SO_FILES[@]}]=$file_name # Append lib to array
-done
-
-# PyTorch-version specific
-# AOTriton dependency only for PyTorch >= 2.4
-if (( $(echo "${PYTORCH_VERSION} 2.4" | awk '{print ($1 >= $2)}') )); then
-    ROCM_SO_FILES+=("libaotriton_v2.so")
-fi
-
-# rocBLAS library files
-ROCBLAS_LIB_SRC=$ROCM_HOME/lib/rocblas/library
-ROCBLAS_LIB_DST=lib/rocblas/library
-ARCH=$(echo $PYTORCH_ROCM_ARCH | sed 's/;/|/g') # Replace ; seperated arch list to bar for grep
-ARCH_SPECIFIC_FILES=$(ls $ROCBLAS_LIB_SRC | grep -E $ARCH)
-OTHER_FILES=$(ls $ROCBLAS_LIB_SRC | grep -v gfx)
-ROCBLAS_LIB_FILES=($ARCH_SPECIFIC_FILES $OTHER_FILES)
-
-# hipblaslt library files
-HIPBLASLT_LIB_SRC=$ROCM_HOME/lib/hipblaslt/library
-HIPBLASLT_LIB_DST=lib/hipblaslt/library
-ARCH_SPECIFIC_FILES=$(ls $HIPBLASLT_LIB_SRC | grep -E $ARCH)
-OTHER_FILES=$(ls $HIPBLASLT_LIB_SRC | grep -v gfx)
-HIPBLASLT_LIB_FILES=($ARCH_SPECIFIC_FILES $OTHER_FILES)
-
-# ROCm library files
-ROCM_SO_PATHS=()
-for lib in "${ROCM_SO_FILES[@]}"
-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
-        echo "Error: Library file $lib is not found." >&2
-        exit 1
-    fi
-    ROCM_SO_PATHS[${#ROCM_SO_PATHS[@]}]="$file_path" # Append lib to array
-done
-
-DEPS_LIST=(
-    ${ROCM_SO_PATHS[*]}
-    ${OS_SO_PATHS[*]}
-)
-
-DEPS_SONAME=(
-    ${ROCM_SO_FILES[*]}
-    ${OS_SO_FILES[*]}
-)
-
-DEPS_AUX_SRCLIST=(
-    "${ROCBLAS_LIB_FILES[@]/#/$ROCBLAS_LIB_SRC/}"
-    "${HIPBLASLT_LIB_FILES[@]/#/$HIPBLASLT_LIB_SRC/}"
-    "/opt/amdgpu/share/libdrm/amdgpu.ids"
-)
-
-DEPS_AUX_DSTLIST=(
-    "${ROCBLAS_LIB_FILES[@]/#/$ROCBLAS_LIB_DST/}"
-    "${HIPBLASLT_LIB_FILES[@]/#/$HIPBLASLT_LIB_DST/}"
-    "share/libdrm/amdgpu.ids"
-)
-
-# MIOpen library files
-MIOPEN_SHARE_SRC=$ROCM_HOME/share/miopen/db
-MIOPEN_SHARE_DST=share/miopen/db
-MIOPEN_SHARE_FILES=($(ls $MIOPEN_SHARE_SRC | grep -E $ARCH))
-DEPS_AUX_SRCLIST+=(${MIOPEN_SHARE_FILES[@]/#/$MIOPEN_SHARE_SRC/})
-DEPS_AUX_DSTLIST+=(${MIOPEN_SHARE_FILES[@]/#/$MIOPEN_SHARE_DST/})
-
-# RCCL library files
-RCCL_SHARE_SRC=$ROCM_HOME/share/rccl/msccl-algorithms
-RCCL_SHARE_DST=share/rccl/msccl-algorithms
-RCCL_SHARE_FILES=($(ls $RCCL_SHARE_SRC))
-DEPS_AUX_SRCLIST+=(${RCCL_SHARE_FILES[@]/#/$RCCL_SHARE_SRC/})
-DEPS_AUX_DSTLIST+=(${RCCL_SHARE_FILES[@]/#/$RCCL_SHARE_DST/})
-
-echo "PYTORCH_ROCM_ARCH: ${PYTORCH_ROCM_ARCH}"
-
-SCRIPTPATH="$( cd "$(dirname "$0")" ; pwd -P )"
-if [[ -z "$BUILD_PYTHONLESS" ]]; then
-    BUILD_SCRIPT=build_common.sh
-else
-    BUILD_SCRIPT=build_libtorch.sh
-fi
-source $SCRIPTPATH/${BUILD_SCRIPT}

.ci/manywheel/test_wheel.sh

@@ -1,26 +0,0 @@
-#!/usr/bin/env bash
-set -e
-
-yum install -y wget git
-
-rm -rf /usr/local/cuda*
-
-# Install Anaconda
-if ! ls /py
-then
-    echo "Miniconda needs to be installed"
-    wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda.sh
-    bash ~/miniconda.sh -b -p /py
-else
-    echo "Miniconda is already installed"
-fi
-
-export PATH="/py/bin:$PATH"
-
-# Anaconda token
-if ls /remote/token
-then
-   source /remote/token
-fi
-
-conda install -y conda-build anaconda-client

.ci/pytorch/build.sh

@@ -49,8 +49,13 @@ if [[ ${BUILD_ENVIRONMENT} == *"parallelnative"* ]]; then
 fi
 
 # Enable LLVM dependency for TensorExpr testing
-export USE_LLVM=/opt/llvm
-export LLVM_DIR=/opt/llvm/lib/cmake/llvm
+if [[ "$BUILD_ENVIRONMENT" == *rocm* ]]; then
+  export USE_LLVM=/opt/rocm/llvm
+  export LLVM_DIR=/opt/rocm/llvm/lib/cmake/llvm
+else
+  export USE_LLVM=/opt/llvm
+  export LLVM_DIR=/opt/llvm/lib/cmake/llvm
+fi
 
 if [[ "$BUILD_ENVIRONMENT" == *executorch* ]]; then
   # To build test_edge_op_registration
@@ -178,7 +183,7 @@ fi
 # sccache will fail for CUDA builds if all cores are used for compiling
 # gcc 7 with sccache seems to have intermittent OOM issue if all cores are used
 if [ -z "$MAX_JOBS" ]; then
-  if { [[ "$BUILD_ENVIRONMENT" == *cuda* ]]; } && which sccache > /dev/null; then
+  if { [[ "$BUILD_ENVIRONMENT" == *cuda* ]] || [[ "$BUILD_ENVIRONMENT" == *gcc7* ]]; } && which sccache > /dev/null; then
     export MAX_JOBS=$(($(nproc) - 1))
   fi
 fi
@@ -203,6 +208,7 @@ if [[ "${BUILD_ENVIRONMENT}" == *clang* ]]; then
 fi
 
 if [[ "$BUILD_ENVIRONMENT" == *-clang*-asan* ]]; then
+  export LDSHARED="clang --shared"
   export USE_CUDA=0
   export USE_ASAN=1
   export UBSAN_FLAGS="-fno-sanitize-recover=all;-fno-sanitize=float-divide-by-zero;-fno-sanitize=float-cast-overflow"
@@ -217,6 +223,10 @@ if [[ "${BUILD_ENVIRONMENT}" == *-pch* ]]; then
     export USE_PRECOMPILED_HEADERS=1
 fi
 
+if [[ "${BUILD_ENVIRONMENT}" == *linux-focal-py3.7-gcc7-build*  ]]; then
+  export USE_GLOO_WITH_OPENSSL=ON
+fi
+
 if [[ "${BUILD_ENVIRONMENT}" != *android* && "${BUILD_ENVIRONMENT}" != *cuda* ]]; then
   export BUILD_STATIC_RUNTIME_BENCHMARK=ON
 fi
@@ -227,7 +237,7 @@ fi
 
 # Do not change workspace permissions for ROCm CI jobs
 # as it can leave workspace with bad permissions for cancelled jobs
-if [[ "$BUILD_ENVIRONMENT" != *rocm* && "$BUILD_ENVIRONMENT" != *s390x* ]]; then
+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() {
@@ -273,7 +283,6 @@ else
     # set only when building other architectures
     # or building non-XLA tests.
     if [[ "$BUILD_ENVIRONMENT" != *rocm*  &&
-          "$BUILD_ENVIRONMENT" != *s390x*   &&
           "$BUILD_ENVIRONMENT" != *xla* ]]; then
       if [[ "$BUILD_ENVIRONMENT" != *py3.8* ]]; then
         # Install numpy-2.0.2 for builds which are backward compatible with 1.X
@@ -336,11 +345,11 @@ else
     CUSTOM_OP_BUILD="${CUSTOM_TEST_ARTIFACT_BUILD_DIR}/custom-op-build"
     CUSTOM_OP_TEST="$PWD/test/custom_operator"
     python --version
-    SITE_PACKAGES="$(python -c 'import site; print(";".join([x for x in site.getsitepackages()] + [x + "/torch" for x in site.getsitepackages()]))')"
+    SITE_PACKAGES="$(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" -DPython_EXECUTABLE="$(which python)" \
+    cmake "$CUSTOM_OP_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch;$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
           -DCMAKE_MODULE_PATH="$CUSTOM_TEST_MODULE_PATH" -DUSE_ROCM="$CUSTOM_TEST_USE_ROCM"
     make VERBOSE=1
     popd
@@ -350,10 +359,10 @@ else
     JIT_HOOK_BUILD="${CUSTOM_TEST_ARTIFACT_BUILD_DIR}/jit-hook-build"
     JIT_HOOK_TEST="$PWD/test/jit_hooks"
     python --version
-    SITE_PACKAGES="$(python -c 'import site; print(";".join([x for x in site.getsitepackages()] + [x + "/torch" for x in site.getsitepackages()]))')"
+    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="$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
+    cmake "$JIT_HOOK_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch;$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
           -DCMAKE_MODULE_PATH="$CUSTOM_TEST_MODULE_PATH" -DUSE_ROCM="$CUSTOM_TEST_USE_ROCM"
     make VERBOSE=1
     popd
@@ -365,7 +374,7 @@ else
     python --version
     mkdir -p "$CUSTOM_BACKEND_BUILD"
     pushd "$CUSTOM_BACKEND_BUILD"
-    cmake "$CUSTOM_BACKEND_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
+    cmake "$CUSTOM_BACKEND_TEST" -DCMAKE_PREFIX_PATH="$SITE_PACKAGES/torch;$SITE_PACKAGES" -DPython_EXECUTABLE="$(which python)" \
           -DCMAKE_MODULE_PATH="$CUSTOM_TEST_MODULE_PATH" -DUSE_ROCM="$CUSTOM_TEST_USE_ROCM"
     make VERBOSE=1
     popd
@@ -398,6 +407,6 @@ fi
 
 # snadampal: skipping it till sccache support added for aarch64
 # https://github.com/pytorch/pytorch/issues/121559
-if [[ "$BUILD_ENVIRONMENT" != *aarch64* &&  "$BUILD_ENVIRONMENT" != *s390x* ]]; then
+if [[ "$BUILD_ENVIRONMENT" != *aarch64* ]]; then
   print_sccache_stats
 fi

.ci/pytorch/common_utils.sh

@@ -191,22 +191,9 @@ function install_torchrec_and_fbgemm() {
   pip_uninstall torchrec-nightly
   pip_uninstall fbgemm-gpu-nightly
   pip_install setuptools-git-versioning scikit-build pyre-extensions
-
-  # TODO (huydhn): I still have no clue on why sccache doesn't work with only fbgemm_gpu here, but it
-  # seems to be an sccache-related issue
-  if [[ "$IS_A100_RUNNER" == "1" ]]; then
-    unset CMAKE_CUDA_COMPILER_LAUNCHER
-    sudo mv /opt/cache/bin /opt/cache/bin-backup
-  fi
-
   # See https://github.com/pytorch/pytorch/issues/106971
   CUDA_PATH=/usr/local/cuda-12.1 pip_install --no-use-pep517 --user "git+https://github.com/pytorch/FBGEMM.git@${fbgemm_commit}#egg=fbgemm-gpu&subdirectory=fbgemm_gpu"
   pip_install --no-use-pep517 --user "git+https://github.com/pytorch/torchrec.git@${torchrec_commit}"
-
-  if [[ "$IS_A100_RUNNER" == "1" ]]; then
-    export CMAKE_CUDA_COMPILER_LAUNCHER=/opt/cache/bin/sccache
-    sudo mv /opt/cache/bin-backup /opt/cache/bin
-  fi
 }
 
 function clone_pytorch_xla() {

.ci/pytorch/create_test_cert.py

@@ -1,4 +1,4 @@
-from datetime import datetime, timedelta, timezone
+from datetime import datetime, timedelta
 from tempfile import mkdtemp
 
 from cryptography import x509
@@ -42,10 +42,11 @@ def create_cert(path, C, ST, L, O, key):
         .issuer_name(issuer)
         .public_key(key.public_key())
         .serial_number(x509.random_serial_number())
-        .not_valid_before(datetime.now(timezone.utc))
+        .not_valid_before(datetime.utcnow())
         .not_valid_after(
             # Our certificate will be valid for 10 days
-            datetime.now(timezone.utc) + timedelta(days=10)
+            datetime.utcnow()
+            + timedelta(days=10)
         )
         .add_extension(
             x509.BasicConstraints(ca=True, path_length=None),
@@ -87,10 +88,11 @@ def sign_certificate_request(path, csr_cert, ca_cert, private_ca_key):
         .issuer_name(ca_cert.subject)
         .public_key(csr_cert.public_key())
         .serial_number(x509.random_serial_number())
-        .not_valid_before(datetime.now(timezone.utc))
+        .not_valid_before(datetime.utcnow())
         .not_valid_after(
             # Our certificate will be valid for 10 days
-            datetime.now(timezone.utc) + timedelta(days=10)
+            datetime.utcnow()
+            + timedelta(days=10)
             # Sign our certificate with our private key
         )
         .sign(private_ca_key, hashes.SHA256())

.ci/pytorch/macos-test.sh

@@ -9,13 +9,15 @@ if [[ -n "$CONDA_ENV" ]]; then
   export PATH="$CONDA_ENV/bin":$PATH
 fi
 
-# Test that OpenMP is enabled
-pushd test
-if [[ ! $(python -c "import torch; print(int(torch.backends.openmp.is_available()))") == "1" ]]; then
-  echo "Build should have OpenMP enabled, but torch.backends.openmp.is_available() is False"
-  exit 1
+# Test that OpenMP is enabled for non-arm64 build
+if [[ ${BUILD_ENVIRONMENT} != *arm64* ]]; then
+  pushd test
+  if [[ ! $(python -c "import torch; print(int(torch.backends.openmp.is_available()))") == "1" ]]; then
+    echo "Build should have OpenMP enabled, but torch.backends.openmp.is_available() is False"
+    exit 1
+  fi
+  popd
 fi
-popd
 
 setup_test_python() {
   # The CircleCI worker hostname doesn't resolve to an address.
@@ -25,9 +27,8 @@ setup_test_python() {
   echo "Ninja version: $(ninja --version)"
   echo "Python version: $(which python) ($(python --version))"
 
-  # Set the limit on open file handles to 16384
-  # might help with intermittent compiler test failures
-  ulimit -n 16384
+  # Increase default limit on open file handles from 256 to 1024
+  ulimit -n 1024
 }
 
 test_python_all() {

.ci/pytorch/test.sh

@@ -49,16 +49,16 @@ NUM_TEST_SHARDS="${NUM_TEST_SHARDS:=1}"
 export VALGRIND=ON
 # export TORCH_INDUCTOR_INSTALL_GXX=ON
 if [[ "$BUILD_ENVIRONMENT" == *clang9* ]]; then
-  # clang9 appears to miscompile code involving std::optional<c10::SymInt>,
+  # clang9 appears to miscompile code involving c10::optional<c10::SymInt>,
   # such that valgrind complains along these lines:
   #
   # Conditional jump or move depends on uninitialised value(s)
   #    at 0x40303A: ~optional_base (Optional.h:281)
   #    by 0x40303A: call (Dispatcher.h:448)
-  #    by 0x40303A: call(at::Tensor const&, c10::ArrayRef<c10::SymInt>, c10::ArrayRef<c10::SymInt>, std::optional<c10::SymInt>) (basic.cpp:10)
+  #    by 0x40303A: call(at::Tensor const&, c10::ArrayRef<c10::SymInt>, c10::ArrayRef<c10::SymInt>, c10::optional<c10::SymInt>) (basic.cpp:10)
   #    by 0x403700: main (basic.cpp:16)
   #  Uninitialised value was created by a stack allocation
-  #    at 0x402AAA: call(at::Tensor const&, c10::ArrayRef<c10::SymInt>, c10::ArrayRef<c10::SymInt>, std::optional<c10::SymInt>) (basic.cpp:6)
+  #    at 0x402AAA: call(at::Tensor const&, c10::ArrayRef<c10::SymInt>, c10::ArrayRef<c10::SymInt>, c10::optional<c10::SymInt>) (basic.cpp:6)
   #
   # The problem does not appear with gcc or newer versions of clang (we tested
   # clang14).  So we suppress valgrind testing for clang9 specifically.
@@ -72,7 +72,7 @@ if [[ "$BUILD_ENVIRONMENT" == *clang9* ]]; then
   #
   # using namespace at;
   #
-  # Tensor call(const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, std::optional<c10::SymInt> storage_offset) {
+  # Tensor call(const at::Tensor & self, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, c10::optional<c10::SymInt> storage_offset) {
   #   auto op = c10::Dispatcher::singleton()
   #       .findSchemaOrThrow(at::_ops::as_strided::name, at::_ops::as_strided::overload_name)
   #       .typed<at::_ops::as_strided::schema>();
@@ -233,8 +233,8 @@ if [[ "$BUILD_ENVIRONMENT" == *asan* ]]; then
     # it depends on a ton of dynamic libraries that most programs aren't gonna
     # have, and it applies to child processes.
 
-    LD_PRELOAD=$(clang --print-file-name=libclang_rt.asan-x86_64.so)
-    export LD_PRELOAD
+    # TODO: get rid of the hardcoded path
+    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
 
@@ -369,27 +369,22 @@ test_inductor_aoti() {
   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
 }
 
-test_inductor_cpp_wrapper() {
-  export TORCHINDUCTOR_CPP_WRAPPER=1
+test_inductor_cpp_wrapper_abi_compatible() {
+  export TORCHINDUCTOR_ABI_COMPATIBLE=1
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
   mkdir -p "$TEST_REPORTS_DIR"
 
-  python benchmarks/dynamo/timm_models.py --device cuda --accuracy --amp \
+  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
+  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"
-
-  python benchmarks/dynamo/torchbench.py --device cuda --accuracy \
-    --bfloat16 --inference --inductor --only hf_T5 --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
-  python benchmarks/dynamo/torchbench.py --device cuda --accuracy \
-    --bfloat16 --inference --inductor --only llama --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
-  python benchmarks/dynamo/torchbench.py --device cuda --accuracy \
-    --bfloat16 --inference --inductor --only moco --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
-  python benchmarks/dynamo/check_accuracy.py \
-    --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" \
-    --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_inference.csv"
 }
 
 # "Global" flags for inductor benchmarking controlled by TEST_CONFIG
@@ -406,10 +401,10 @@ pr_time_benchmarks() {
 
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
   mkdir -p "$TEST_REPORTS_DIR"
-  PYTHONPATH=$(pwd)/benchmarks/dynamo/pr_time_benchmarks source benchmarks/dynamo/pr_time_benchmarks/benchmark_runner.sh "$TEST_REPORTS_DIR/pr_time_benchmarks_results.csv" "benchmarks/dynamo/pr_time_benchmarks/benchmarks"
+  PYTHONPATH=$(pwd)/benchmarks/dynamo/pr_time_benchmarks source benchmarks/dynamo/pr_time_benchmarks/benchmark_runner.sh "$TEST_REPORTS_DIR/pr_time_benchmarks_after.txt" "benchmarks/dynamo/pr_time_benchmarks/benchmarks"
   echo "benchmark results on current PR: "
-  cat  "$TEST_REPORTS_DIR/pr_time_benchmarks_results.csv"
-  PYTHONPATH=$(pwd)/benchmarks/dynamo/pr_time_benchmarks python benchmarks/dynamo/pr_time_benchmarks/check_results.py "benchmarks/dynamo/pr_time_benchmarks/expected_results.csv" "$TEST_REPORTS_DIR/pr_time_benchmarks_results.csv" "$TEST_REPORTS_DIR/new_expected_results.csv"
+  cat  "$TEST_REPORTS_DIR/pr_time_benchmarks_after.txt"
+
 }
 
 if [[ "${TEST_CONFIG}" == *pr_time_benchmarks* ]]; then
@@ -517,7 +512,7 @@ test_perf_for_dashboard() {
               "${target_flag[@]}" --"$mode" --"$dtype" --export --disable-cudagraphs "$@" \
               --output "$TEST_REPORTS_DIR/${backend}_export_${suite}_${dtype}_${mode}_${device}_${target}.csv"
         fi
-        $TASKSET python "benchmarks/dynamo/$suite.py" \
+        TORCHINDUCTOR_ABI_COMPATIBLE=1 $TASKSET python "benchmarks/dynamo/$suite.py" \
             "${target_flag[@]}" --"$mode" --"$dtype" --export-aot-inductor --disable-cudagraphs "$@" \
             --output "$TEST_REPORTS_DIR/${backend}_aot_inductor_${suite}_${dtype}_${mode}_${device}_${target}.csv"
       fi
@@ -572,6 +567,13 @@ test_single_dynamo_benchmark() {
     test_perf_for_dashboard "$suite" \
       "${DYNAMO_BENCHMARK_FLAGS[@]}" "$@" "${partition_flags[@]}"
   else
+    if [[ "${TEST_CONFIG}" == *aot_inductor* && "${TEST_CONFIG}" != *cpu_aot_inductor* ]]; then
+      # Test AOTInductor with the ABI-compatible mode on CI
+      # This can be removed once the ABI-compatible mode becomes default.
+      # For CPU device, we perfer non ABI-compatible mode on CI when testing AOTInductor.
+      export TORCHINDUCTOR_ABI_COMPATIBLE=1
+    fi
+
     if [[ "${TEST_CONFIG}" == *_avx2* ]]; then
       TEST_CONFIG=${TEST_CONFIG//_avx2/}
     fi
@@ -605,11 +607,6 @@ test_inductor_halide() {
   assert_git_not_dirty
 }
 
-test_inductor_triton_cpu() {
-  python test/run_test.py --include inductor/test_triton_cpu_backend.py --verbose
-  assert_git_not_dirty
-}
-
 test_dynamo_benchmark() {
   # Usage: test_dynamo_benchmark huggingface 0
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
@@ -647,12 +644,32 @@ test_inductor_torchbench_smoketest_perf() {
   TEST_REPORTS_DIR=$(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 "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv"
+  python benchmarks/dynamo/check_accuracy.py \
+    --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" \
+    --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_inference.csv"
+
   python benchmarks/dynamo/torchbench.py --device cuda --performance --backend inductor --float16 --training \
     --batch-size-file "$(realpath benchmarks/dynamo/torchbench_models_list.txt)" --only hf_Bert \
     --output "$TEST_REPORTS_DIR/inductor_training_smoketest.csv"
   # The threshold value needs to be actively maintained to make this check useful
   python benchmarks/dynamo/check_perf_csv.py -f "$TEST_REPORTS_DIR/inductor_training_smoketest.csv" -t 1.4
 
+  TORCHINDUCTOR_ABI_COMPATIBLE=1 python benchmarks/dynamo/torchbench.py --device cuda --performance --bfloat16 --inference \
+    --export-aot-inductor --only nanogpt --output "$TEST_REPORTS_DIR/inductor_inference_smoketest.csv"
+  # The threshold value needs to be actively maintained to make this check useful
+  # The perf number of nanogpt seems not very stable, e.g.
+  # https://github.com/pytorch/pytorch/actions/runs/7158691360/job/19491437314,
+  # and thus we lower its threshold to reduce flakiness. If this continues to be a problem,
+  # we switch to use some other model.
+  python benchmarks/dynamo/check_perf_csv.py -f "$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
     python benchmarks/dynamo/torchbench.py --device cuda --performance --backend inductor --amp --training \
@@ -696,10 +713,6 @@ test_inductor_set_cpu_affinity(){
     export KMP_BLOCKTIME=1
   fi
   cores=$(test_inductor_get_core_number)
-  # Set number of cores to 16 on Aarch64 for performance runs.
-  if [[ "${TEST_CONFIG}" == *aarch64* && $cores -gt 16 ]]; then
-    cores=16
-  fi
   export OMP_NUM_THREADS=$cores
   end_core=$((cores-1))
   export TASKSET="taskset -c 0-$end_core"
@@ -736,9 +749,19 @@ test_inductor_torchbench_cpu_smoketest_perf(){
     fi
     cat "$output_name"
     # The threshold value needs to be actively maintained to make this check useful.
-    # Allow 1% variance for CPU perf to accommodate perf fluctuation
-    python benchmarks/dynamo/check_perf_csv.py -f "$output_name" -t "$speedup_target" -s 0.99
+    python benchmarks/dynamo/check_perf_csv.py -f "$output_name" -t "$speedup_target"
   done
+
+  # Add a few ABI-compatible accuracy tests for CPU. These can be removed once we turn on ABI-compatible as default.
+  TORCHINDUCTOR_ABI_COMPATIBLE=1 python benchmarks/dynamo/timm_models.py --device cpu --accuracy \
+    --bfloat16 --inference --export-aot-inductor --disable-cudagraphs --only adv_inception_v3 \
+    --output "$TEST_REPORTS_DIR/aot_inductor_smoke_test.csv"
+  TORCHINDUCTOR_ABI_COMPATIBLE=1 python benchmarks/dynamo/timm_models.py --device cpu --accuracy \
+    --bfloat16 --inference --export-aot-inductor --disable-cudagraphs --only beit_base_patch16_224 \
+    --output "$TEST_REPORTS_DIR/aot_inductor_smoke_test.csv"
+  python benchmarks/dynamo/check_accuracy.py \
+    --actual "$TEST_REPORTS_DIR/aot_inductor_smoke_test.csv" \
+    --expected "benchmarks/dynamo/ci_expected_accuracy/aot_inductor_timm_inference.csv"
 }
 
 test_torchbench_gcp_smoketest(){
@@ -1360,16 +1383,14 @@ test_executorch() {
   assert_git_not_dirty
 }
 
-test_linux_aarch64() {
+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 \
-        --shard "$SHARD_NUMBER" "$NUM_TEST_SHARDS" --verbose
+       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 \
-       --shard "$SHARD_NUMBER" "$NUM_TEST_SHARDS" --verbose
+       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 \
@@ -1379,8 +1400,7 @@ test_linux_aarch64() {
        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 inductor/test_memory \
-       --shard "$SHARD_NUMBER" "$NUM_TEST_SHARDS" --verbose
+       inductor/test_torchinductor_codegen_dynamic_shapes inductor/test_torchinductor_dynamic_shapes --verbose
 }
 
 if ! [[ "${BUILD_ENVIRONMENT}" == *libtorch* || "${BUILD_ENVIRONMENT}" == *-bazel-* ]]; then
@@ -1413,8 +1433,6 @@ elif [[ "${TEST_CONFIG}" == *inductor_distributed* ]]; then
   test_inductor_distributed
 elif [[ "${TEST_CONFIG}" == *inductor-halide* ]]; then
   test_inductor_halide
-elif [[ "${TEST_CONFIG}" == *inductor-triton-cpu* ]]; then
-  test_inductor_triton_cpu
 elif [[ "${TEST_CONFIG}" == *inductor-micro-benchmark* ]]; then
   test_inductor_micro_benchmark
 elif [[ "${TEST_CONFIG}" == *huggingface* ]]; then
@@ -1431,13 +1449,14 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
   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=$((SHARD_NUMBER-1))
   # https://github.com/opencv/opencv-python/issues/885
   pip_install opencv-python==4.8.0.74
   if [[ "${TEST_CONFIG}" == *inductor_torchbench_smoketest_perf* ]]; then
-    checkout_install_torchbench hf_Bert hf_Albert timm_vision_transformer
+    checkout_install_torchbench hf_Bert hf_Albert nanogpt timm_vision_transformer
     PYTHONPATH=$(pwd)/torchbench test_inductor_torchbench_smoketest_perf
   elif [[ "${TEST_CONFIG}" == *inductor_torchbench_cpu_smoketest_perf* ]]; then
     checkout_install_torchbench timm_vision_transformer phlippe_densenet basic_gnn_edgecnn \
@@ -1456,11 +1475,9 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
     fi
     PYTHONPATH=$(pwd)/torchbench test_dynamo_benchmark torchbench "$id"
   fi
-elif [[ "${TEST_CONFIG}" == *inductor_cpp_wrapper* ]]; then
-  install_torchaudio cuda
+elif [[ "${TEST_CONFIG}" == *inductor_cpp_wrapper_abi_compatible* ]]; then
   install_torchvision
-  checkout_install_torchbench hf_T5 llama moco
-  PYTHONPATH=$(pwd)/torchbench test_inductor_cpp_wrapper
+  test_inductor_cpp_wrapper_abi_compatible
 elif [[ "${TEST_CONFIG}" == *inductor* ]]; then
   install_torchvision
   test_inductor_shard "${SHARD_NUMBER}"

.ci/pytorch/win-build.sh

@@ -26,7 +26,7 @@ fi
 export SCRIPT_HELPERS_DIR=$SCRIPT_PARENT_DIR/win-test-helpers
 
 set +ex
-grep -E -R 'PyLong_(From|As)(Unsigned|)Long\(' --exclude=python_numbers.h  --exclude=pythoncapi_compat.h --exclude=eval_frame.c torch/
+grep -E -R 'PyLong_(From|As)(Unsigned|)Long\(' --exclude=python_numbers.h --exclude=eval_frame.c torch/
 PYLONG_API_CHECK=$?
 if [[ $PYLONG_API_CHECK == 0 ]]; then
   echo "Usage of PyLong_{From,As}{Unsigned}Long API may lead to overflow errors on Windows"

.ci/pytorch/win-test.sh

@@ -46,9 +46,6 @@ python -m pip install tlparse==0.3.25
 # Install parameterized
 python -m pip install parameterized==0.8.1
 
-# Install pulp for testing ilps under torch\distributed\_tools
-python -m pip install pulp==2.9.0
-
 run_tests() {
     # Run nvidia-smi if available
     for path in '/c/Program Files/NVIDIA Corporation/NVSMI/nvidia-smi.exe' /c/Windows/System32/nvidia-smi.exe; do

.circleci/scripts/binary_linux_test.sh

@@ -27,11 +27,12 @@ if [[ "$PACKAGE_TYPE" == conda ]]; then
   source activate testenv >/dev/null
 elif [[ "$PACKAGE_TYPE" != libtorch ]]; then
   python_path="/opt/python/cp\$python_nodot-cp\${python_nodot}"
-  if [[ "\$python_nodot" = *t ]]; then
-    python_digits="\$(echo $DESIRED_PYTHON | tr -cd [:digit:])"
-    python_path="/opt/python/cp\$python_digits-cp\${python_digits}t"
+  # Prior to Python 3.8 paths were suffixed with an 'm'
+  if [[ -d  "\${python_path}/bin" ]]; then
+    export PATH="\${python_path}/bin:\$PATH"
+  elif [[ -d "\${python_path}m/bin" ]]; then
+    export PATH="\${python_path}m/bin:\$PATH"
   fi
-  export PATH="\${python_path}/bin:\$PATH"
 fi
 
 EXTRA_CONDA_FLAGS=""

.clang-format

@@ -44,9 +44,7 @@ ContinuationIndentWidth: 4
 Cpp11BracedListStyle: true
 DerivePointerAlignment: false
 DisableFormat:   false
-ForEachMacros:
-  - FOR_EACH_RANGE
-  - FOR_EACH
+ForEachMacros:   [ FOR_EACH_RANGE, FOR_EACH, ]
 IncludeCategories:
   - Regex:           '^<.*\.h(pp)?>'
     Priority:        1
@@ -60,24 +58,6 @@ IndentWrappedFunctionNames: false
 KeepEmptyLinesAtTheStartOfBlocks: false
 MacroBlockBegin: ''
 MacroBlockEnd:   ''
-Macros:
-  - >-
-    PyObject_HEAD_INIT(type)={
-        /* this is not exactly match with PyObject_HEAD_INIT in Python source code
-         * but it is enough for clang-format */
-        { 0xFFFFFFFF },
-        (type)
-    },
-  - >-
-    PyVarObject_HEAD_INIT(type, size)={
-        {
-            /* manually expand PyObject_HEAD_INIT(type) above
-             * because clang-format do not support recursive expansion */
-            { 0xFFFFFFFF },
-            (type)
-        },
-        (size)
-    },
 MaxEmptyLinesToKeep: 1
 NamespaceIndentation: None
 PenaltyBreakBeforeFirstCallParameter: 1
@@ -99,11 +79,7 @@ SpacesInContainerLiterals: true
 SpacesInCStyleCastParentheses: false
 SpacesInParentheses: false
 SpacesInSquareBrackets: false
-Standard:        c++17
-StatementMacros:
-  - PyObject_HEAD
-  - PyObject_VAR_HEAD
-  - PyException_HEAD
+Standard:        Cpp11
 TabWidth:        8
 UseTab:          Never
 ---

.github/ISSUE_TEMPLATE.md

@@ -0,0 +1,38 @@
+If you have a question or would like help and support, please ask at our
+[forums](https://discuss.pytorch.org/).
+
+If you are submitting a feature request, please preface the title with [feature request].
+If you are submitting a bug report, please fill in the following details.
+
+## Issue description
+
+Provide a short description.
+
+## Code example
+
+Please try to provide a minimal example to repro the bug.
+Error messages and stack traces are also helpful.
+
+## System Info
+Please copy and paste the output from our
+[environment collection script](https://raw.githubusercontent.com/pytorch/pytorch/main/torch/utils/collect_env.py)
+(or fill out the checklist below manually).
+
+You can get the script and run it with:
+```
+wget https://raw.githubusercontent.com/pytorch/pytorch/main/torch/utils/collect_env.py
+# For security purposes, please check the contents of collect_env.py before running it.
+python collect_env.py
+```
+
+- PyTorch or Caffe2:
+- How you installed PyTorch (conda, pip, source):
+- Build command you used (if compiling from source):
+- OS:
+- PyTorch version:
+- Python version:
+- CUDA/cuDNN version:
+- GPU models and configuration:
+- GCC version (if compiling from source):
+- CMake version:
+- Versions of any other relevant libraries:

.github/ISSUE_TEMPLATE/ci-sev.md

@@ -5,8 +5,7 @@ about: Tracking incidents for PyTorch's CI infra.
 
 > NOTE: Remember to label this issue with "`ci: sev`"
 
- <!-- uncomment the below line if you don't want this SEV to block merges -->
- <!--  **MERGE BLOCKING** -->
+**MERGE BLOCKING** <!-- remove this line if you don't want this SEV to block merges -->
 
 ## Current Status
 *Status could be: preemptive, ongoing, mitigated, closed. Also tell people if they need to take action to fix it (i.e. rebase)*.

.github/actionlint.yaml

@@ -32,6 +32,30 @@ self-hosted-runner:
     - lf.linux.8xlarge.nvidia.gpu
     - lf.linux.16xlarge.nvidia.gpu
     - lf.linux.g5.4xlarge.nvidia.gpu
+    # Organization-wide AWS Linux Runners with new Amazon 2023 AMI
+    - amz2023.linux.large
+    - amz2023.linux.2xlarge
+    - amz2023.linux.4xlarge
+    - amz2023.linux.12xlarge
+    - amz2023.linux.24xlarge
+    - amz2023.linux.arm64.2xlarge
+    - amz2023.linux.arm64.m7g.4xlarge
+    - amz2023.linux.arm64.m7g.4xlarge.ephemeral
+    - amz2023.linux.4xlarge.nvidia.gpu
+    - amz2023.linux.8xlarge.nvidia.gpu
+    - amz2023.linux.16xlarge.nvidia.gpu
+    - amz2023.linux.g5.4xlarge.nvidia.gpu
+    # Pytorch/pytorch AWS Linux Runners with the new Amazon 2023 AMI on Linux Foundation account
+    - amz2023.lf.linux.large
+    - amz2023.lf.linux.2xlarge
+    - amz2023.lf.linux.4xlarge
+    - amz2023.lf.linux.12xlarge
+    - amz2023.lf.linux.24xlarge
+    - amz2023.lf.linux.arm64.2xlarge
+    - amz2023.lf.linux.4xlarge.nvidia.gpu
+    - amz2023.lf.linux.8xlarge.nvidia.gpu
+    - amz2023.lf.linux.16xlarge.nvidia.gpu
+    - amz2023.lf.linux.g5.4xlarge.nvidia.gpu
     # Repo-specific IBM hosted S390x runner
     - linux.s390x
     # Organization wide AWS Windows runners

.github/actions/checkout-pytorch/action.yml

@@ -18,14 +18,8 @@ inputs:
 runs:
   using: composite
   steps:
-    - name: Check if in a container runner
-      shell: bash
-      id: check_container_runner
-      run: echo "IN_CONTAINER_RUNNER=$(if [ -f /.inarc ] || [ -f /.incontainer ]; then echo true ; else echo false; fi)" >> "$GITHUB_OUTPUT"
-
     - name: Clean workspace
       shell: bash
-      if: ${{ steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
       env:
         NO_SUDO: ${{ inputs.no-sudo }}
       run: |

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

@@ -85,25 +85,15 @@ runs:
       with:
         docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
-    - name: Check if in a container runner
+    - name: Check if in a ARC runner
       shell: bash
-      id: check_container_runner
-      run: echo "IN_CONTAINER_RUNNER=$(if [ -f /.inarc ] || [ -f /.incontainer ]; then echo true ; else echo false; fi)" >> "$GITHUB_OUTPUT"
+      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: ${{ contains(inputs.build-environment, 'cuda') && !contains(matrix.config, 'nogpu') && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
-
-    - name: Setup GPU_FLAG for docker run
-      id: setup-gpu-flag
-      run: echo "GPU_FLAG=--gpus all -e NVIDIA_DRIVER_CAPABILITIES=all" >> "${GITHUB_ENV}"
-      if: ${{ contains(inputs.build-environment, 'cuda') && !contains(matrix.config, 'nogpu') && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'true' }}
-
-    - name: Setup SCCACHE_SERVER_PORT environment for docker run when on container
-      id: setup-sscache-port-flag
-      run: echo "SCCACHE_SERVER_PORT_DOCKER_FLAG=-e SCCACHE_SERVER_PORT=$((RUNNER_UID + 4226))" >> "${GITHUB_ENV}"
-      if: ${{ steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'true' }}
+      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
@@ -111,7 +101,7 @@ runs:
         sudo nvidia-smi -pm 1
         sudo nvidia-smi -ac 1215,1410
         nvidia-smi
-      if: ${{ contains(matrix.runner, 'a100') && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
+      if: contains(matrix.runner, 'a100')
 
     - name: Start monitoring script
       id: monitor-script
@@ -182,7 +172,6 @@ runs:
         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_REGION: us-east-1
         SCCACHE_S3_KEY_PREFIX: ${{ github.workflow }}
         SHM_SIZE: ${{ contains(inputs.build-environment, 'cuda') && '2g' || '1g' }}
         DOCKER_IMAGE: ${{ inputs.docker-image }}
@@ -192,9 +181,6 @@ runs:
         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 }}
-        SCRIBE_GRAPHQL_ACCESS_TOKEN: ${{ secrets.SCRIBE_GRAPHQL_ACCESS_TOKEN }}
-        IS_A100_RUNNER: ${{ contains(matrix.runner, 'a100') && '1' || '0' }}
-
       shell: bash
       run: |
         set -x
@@ -213,7 +199,6 @@ runs:
         # shellcheck disable=SC2086,SC2090
         container_name=$(docker run \
           ${GPU_FLAG:-} \
-          ${SCCACHE_SERVER_PORT_DOCKER_FLAG:-} \
           -e BUILD_ENVIRONMENT \
           -e PR_NUMBER \
           -e GITHUB_ACTIONS \
@@ -242,7 +227,6 @@ runs:
           -e PR_LABELS \
           -e MAX_JOBS="$(nproc --ignore=2)" \
           -e SCCACHE_BUCKET \
-          -e SCCACHE_REGION \
           -e SCCACHE_S3_KEY_PREFIX \
           -e XLA_CUDA \
           -e XLA_CLANG_CACHE_S3_BUCKET_NAME \
@@ -250,9 +234,7 @@ runs:
           -e PYTORCH_TEST_RERUN_DISABLED_TESTS \
           -e SKIP_SCCACHE_INITIALIZATION=1 \
           -e HUGGING_FACE_HUB_TOKEN \
-          -e SCRIBE_GRAPHQL_ACCESS_TOKEN \
           -e DASHBOARD_TAG \
-          -e IS_A100_RUNNER \
           --env-file="/tmp/github_env_${GITHUB_RUN_ID}" \
           --security-opt seccomp=unconfined \
           --cap-add=SYS_PTRACE \
@@ -323,7 +305,7 @@ runs:
 
     - name: Teardown Linux
       uses: pytorch/test-infra/.github/actions/teardown-linux@main
-      if: always() && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false'
+      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

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

@@ -28,14 +28,14 @@ runs:
         echo "instance-type: $(get_ec2_metadata instance-type)"
         echo "system info $(uname -a)"
 
-    - name: Check if in a container runner
+    - name: Check if in a ARC runner
       shell: bash
-      id: check_container_runner
-      run: echo "IN_CONTAINER_RUNNER=$(if [ -f /.inarc ] || [ -f /.incontainer ]; then echo true ; else echo false; fi)" >> "$GITHUB_OUTPUT"
+      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_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
+      if: ${{ steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
       run: |
         if systemctl is-active --quiet docker; then
             echo "Docker daemon is running...";
@@ -73,7 +73,7 @@ runs:
         env | grep '^CI' >> "/tmp/github_env_${GITHUB_RUN_ID}"
 
     - name: Kill any existing containers, clean up images
-      if: ${{ steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
+      if: ${{ steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
       shell: bash
       run: |
         # ignore expansion of "docker ps -q" since it could be empty
@@ -116,7 +116,7 @@ runs:
     - name: Check that the docker daemon is running
       shell: bash
       continue-on-error: true
-      if: ${{ steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'true' }}
+      if: ${{ steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'true' }}
       run: |
         set +x
 

.github/ci_commit_pins/audio.txt

@@ -1 +1 @@
-3f0569939c4369bec943fc27d1c9d8dfbc828c26
+97ed7b36b7a741253d4e41e4da3c901d83294503

.github/labeler.yml

@@ -98,9 +98,3 @@
 "module: distributed_checkpoint":
 - torch/distributed/checkpoint/**
 - test/distributed/checkpoint/**
-
-"module: compiled autograd":
-- torch/csrc/dynamo/python_compiled_autograd.cpp
-- torch/csrc/dynamo/compiled_autograd.h
-- torch/_dynamo/compiled_autograd.py
-- torch/inductor/test_compiled_autograd.py

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

@@ -7,14 +7,10 @@
 #   runners. Runners listed here will be available as self hosted
 #   runners, configuration is directly pulled from the main branch.
 #
+# NOTE (Apr, 5, 2021): Linux runners are currently all an amazonlinux2
 #
-# NOTES:
-#  - Linux runners are by default non-ephemeral to reduce the amount of CreateInstaces calls
-#    to avoid RequestLimitExceeded issues
-#  - When updating this file, run the following command to validate the YAML and to generate
-#    corresponding versions of scale-config for the pytorch/pytorch repo and merge the
-#    pytorch/pytorch changes before merging these changes.
-#    `python .github/scripts/validate_scale_config.py --test-infra-repo-root [path_to_test-infra_root] --pytorch-repo-root [path_to_pytorch_root]``
+# NOTE (Jan 5, 2021): Linux runners are all non-ephemeral to reduce the amount of CreateInstaces calls
+#                     to avoid RequestLimitExceeded issues
 #
 # TODO: Add some documentation on how the auto-scaling works
 #
@@ -33,38 +29,60 @@ runner_types:
     disk_size: 200
     instance_type: c5.12xlarge
     is_ephemeral: false
-    max_available: 2000
+    max_available: 1000
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.10xlarge.avx2:
     disk_size: 200
     instance_type: m4.10xlarge
     is_ephemeral: false
     max_available: 450
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.24xl.spr-metal:
     disk_size: 200
     instance_type: c7i.metal-24xl
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.16xlarge.spr:
     disk_size: 200
     instance_type: c7i.16xlarge
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.9xlarge.ephemeral:
     disk_size: 200
     instance_type: c5.9xlarge
     is_ephemeral: true
     max_available: 50
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
     variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
       am2:
         ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.12xlarge.ephemeral:
@@ -73,140 +91,240 @@ runner_types:
     is_ephemeral: true
     max_available: 300
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.16xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g3.16xlarge
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.24xlarge:
     disk_size: 150
     instance_type: c5.24xlarge
     is_ephemeral: false
     max_available: 500
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.24xlarge.ephemeral:
     disk_size: 150
     instance_type: c5.24xlarge
     is_ephemeral: true
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.2xlarge:
     disk_size: 150
     instance_type: c5.2xlarge
     is_ephemeral: false
     max_available: 3120
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.4xlarge:
     disk_size: 150
     instance_type: c5.4xlarge
     is_ephemeral: false
     max_available: 1000
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.4xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g3.4xlarge
     is_ephemeral: false
     max_available: 1000
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.8xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g3.8xlarge
     is_ephemeral: false
     max_available: 400
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.g4dn.12xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g4dn.12xlarge
     is_ephemeral: false
     max_available: 250
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.g4dn.metal.nvidia.gpu:
     disk_size: 150
     instance_type: g4dn.metal
     is_ephemeral: false
     max_available: 300
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.g5.48xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g5.48xlarge
     is_ephemeral: false
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.g5.12xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g5.12xlarge
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.g5.4xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g5.4xlarge
     is_ephemeral: false
     max_available: 2400
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.g6.4xlarge.experimental.nvidia.gpu:
     disk_size: 150
     instance_type: g6.4xlarge
     is_ephemeral: false
     max_available: 50
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.large:
     max_available: 1200
     disk_size: 15
     instance_type: c5.large
     is_ephemeral: false
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.c.linux.arm64.2xlarge:
     disk_size: 256
     instance_type: t4g.2xlarge
     is_ephemeral: false
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.c.linux.arm64.m7g.4xlarge:
     disk_size: 256
     instance_type: m7g.4xlarge
     is_ephemeral: false
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.c.linux.arm64.2xlarge.ephemeral:
     disk_size: 256
     instance_type: t4g.2xlarge
     is_ephemeral: true
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.c.linux.arm64.m7g.4xlarge.ephemeral:
     disk_size: 256
     instance_type: m7g.4xlarge
     is_ephemeral: true
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.c.linux.arm64.m7g.metal:
     disk_size: 256
     instance_type: m7g.metal
     is_ephemeral: false
     max_available: 100
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.c.windows.g4dn.xlarge:
     disk_size: 256
     instance_type: g4dn.xlarge
@@ -241,7 +359,7 @@ runner_types:
     disk_size: 256
     instance_type: p3.2xlarge
     is_ephemeral: false
-    max_available: 300
+    max_available: 150
     os: windows
   lf.c.windows.g5.4xlarge.nvidia.gpu:
     disk_size: 256

.github/lf-scale-config.yml

@@ -7,14 +7,10 @@
 #   runners. Runners listed here will be available as self hosted
 #   runners, configuration is directly pulled from the main branch.
 #
+# NOTE (Apr, 5, 2021): Linux runners are currently all an amazonlinux2
 #
-# NOTES:
-#  - Linux runners are by default non-ephemeral to reduce the amount of CreateInstaces calls
-#    to avoid RequestLimitExceeded issues
-#  - When updating this file, run the following command to validate the YAML and to generate
-#    corresponding versions of scale-config for the pytorch/pytorch repo and merge the
-#    pytorch/pytorch changes before merging these changes.
-#    `python .github/scripts/validate_scale_config.py --test-infra-repo-root [path_to_test-infra_root] --pytorch-repo-root [path_to_pytorch_root]``
+# NOTE (Jan 5, 2021): Linux runners are all non-ephemeral to reduce the amount of CreateInstaces calls
+#                     to avoid RequestLimitExceeded issues
 #
 # TODO: Add some documentation on how the auto-scaling works
 #
@@ -33,38 +29,60 @@ runner_types:
     disk_size: 200
     instance_type: c5.12xlarge
     is_ephemeral: false
-    max_available: 2000
+    max_available: 1000
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.10xlarge.avx2:
     disk_size: 200
     instance_type: m4.10xlarge
     is_ephemeral: false
     max_available: 450
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.24xl.spr-metal:
     disk_size: 200
     instance_type: c7i.metal-24xl
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.16xlarge.spr:
     disk_size: 200
     instance_type: c7i.16xlarge
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.9xlarge.ephemeral:
     disk_size: 200
     instance_type: c5.9xlarge
     is_ephemeral: true
     max_available: 50
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
     variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
       am2:
         ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.12xlarge.ephemeral:
@@ -73,140 +91,240 @@ runner_types:
     is_ephemeral: true
     max_available: 300
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.16xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g3.16xlarge
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.24xlarge:
     disk_size: 150
     instance_type: c5.24xlarge
     is_ephemeral: false
     max_available: 500
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.24xlarge.ephemeral:
     disk_size: 150
     instance_type: c5.24xlarge
     is_ephemeral: true
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.2xlarge:
     disk_size: 150
     instance_type: c5.2xlarge
     is_ephemeral: false
     max_available: 3120
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.4xlarge:
     disk_size: 150
     instance_type: c5.4xlarge
     is_ephemeral: false
     max_available: 1000
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.4xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g3.4xlarge
     is_ephemeral: false
     max_available: 1000
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.8xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g3.8xlarge
     is_ephemeral: false
     max_available: 400
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.g4dn.12xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g4dn.12xlarge
     is_ephemeral: false
     max_available: 250
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.g4dn.metal.nvidia.gpu:
     disk_size: 150
     instance_type: g4dn.metal
     is_ephemeral: false
     max_available: 300
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.g5.48xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g5.48xlarge
     is_ephemeral: false
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.g5.12xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g5.12xlarge
     is_ephemeral: false
     max_available: 150
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.g5.4xlarge.nvidia.gpu:
     disk_size: 150
     instance_type: g5.4xlarge
     is_ephemeral: false
     max_available: 2400
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.g6.4xlarge.experimental.nvidia.gpu:
     disk_size: 150
     instance_type: g6.4xlarge
     is_ephemeral: false
     max_available: 50
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.large:
     max_available: 1200
     disk_size: 15
     instance_type: c5.large
     is_ephemeral: false
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-x86_64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-x86_64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-x86_64-ebs
   lf.linux.arm64.2xlarge:
     disk_size: 256
     instance_type: t4g.2xlarge
     is_ephemeral: false
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.linux.arm64.m7g.4xlarge:
     disk_size: 256
     instance_type: m7g.4xlarge
     is_ephemeral: false
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.linux.arm64.2xlarge.ephemeral:
     disk_size: 256
     instance_type: t4g.2xlarge
     is_ephemeral: true
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.linux.arm64.m7g.4xlarge.ephemeral:
     disk_size: 256
     instance_type: m7g.4xlarge
     is_ephemeral: true
     max_available: 200
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.linux.arm64.m7g.metal:
     disk_size: 256
     instance_type: m7g.metal
     is_ephemeral: false
     max_available: 100
     os: linux
-    ami: al2023-ami-2023.5.202*-kernel-6.1-arm64
+    ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+    variants:
+      amz2023:
+        ami: al2023-ami-2023.5.20240701.0-kernel-6.1-arm64
+      am2:
+        ami: amzn2-ami-hvm-2.0.20240306.2-arm64-gp2
   lf.windows.g4dn.xlarge:
     disk_size: 256
     instance_type: g4dn.xlarge
@@ -241,7 +359,7 @@ runner_types:
     disk_size: 256
     instance_type: p3.2xlarge
     is_ephemeral: false
-    max_available: 300
+    max_available: 150
     os: windows
   lf.windows.g5.4xlarge.nvidia.gpu:
     disk_size: 256

.github/merge_rules.yaml

@@ -544,7 +544,6 @@
   - anijain2305
   - bdhirsh
   - zou3519
-  - isuruf
   mandatory_checks_name:
   - EasyCLA
   - Lint

.github/pytorch-probot.yml

@@ -16,13 +16,11 @@ ciflow_push_tags:
 - ciflow/nightly
 - ciflow/periodic
 - ciflow/rocm
-- ciflow/s390
 - ciflow/slow
 - ciflow/trunk
 - ciflow/unstable
 - ciflow/xpu
 - ciflow/torchbench
-- ciflow/autoformat
 retryable_workflows:
 - pull
 - trunk

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

@@ -1,4 +1,4 @@
 # iOS simulator requirements
 coremltools==5.0b5
 protobuf==3.20.2
-optree==0.13.0
+optree==0.12.1

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

@@ -27,7 +27,7 @@ pytest-cpp==2.3.0
 rockset==1.0.3
 z3-solver==4.12.2.0
 tensorboard==2.13.0
-optree==0.13.0
+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

.github/scripts/generate_binary_build_matrix.py

@@ -77,7 +77,6 @@ PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {
         "nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'"
@@ -334,7 +333,7 @@ def generate_wheels_matrix(
         package_type = "manywheel"
 
     if python_versions is None:
-        python_versions = FULL_PYTHON_VERSIONS + ["3.13", "3.13t"]
+        python_versions = FULL_PYTHON_VERSIONS + ["3.13"]
 
     if arches is None:
         # Define default compute archivectures
@@ -369,15 +368,8 @@ def generate_wheels_matrix(
 
             # TODO: Enable python 3.13 on rocm, aarch64, windows
             if (
-                gpu_arch_type == "rocm"
-                or os not in ["linux", "linux-s390x", "macos-arm64"]
-            ) and python_version in ["3.13", "3.13t"]:
-                continue
-
-            # TODO: Enable python 3.13t on xpu and cpu-s390x or MacOS
-            if (
-                gpu_arch_type in ["xpu", "cpu-s390x"] or os == "macos-arm64"
-            ) and python_version == "3.13t":
+                gpu_arch_type == "rocm" or (os != "linux" and os != "linux-s390x")
+            ) and python_version == "3.13":
                 continue
 
             if use_split_build and (
@@ -411,7 +403,7 @@ def generate_wheels_matrix(
                         "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
                         "package_type": package_type,
                         "pytorch_extra_install_requirements": (
-                            PYTORCH_EXTRA_INSTALL_REQUIREMENTS[arch_version]
+                            PYTORCH_EXTRA_INSTALL_REQUIREMENTS[arch_version]  # fmt: skip
                             if os != "linux-aarch64"
                             else ""
                         ),
@@ -459,7 +451,7 @@ def generate_wheels_matrix(
                             ".", "_"
                         ),
                         "pytorch_extra_install_requirements": (
-                            PYTORCH_EXTRA_INSTALL_REQUIREMENTS["12.1"]
+                            PYTORCH_EXTRA_INSTALL_REQUIREMENTS["12.1"]  # fmt: skip
                             if os != "linux" and gpu_arch_type != "xpu"
                             else ""
                         ),

.github/scripts/lintrunner.sh

@@ -17,11 +17,6 @@ if [[ -d "${CACHE_DIRECTORY}" ]]; then
     cp -r "${CACHE_DIRECTORY}" . || true
 fi
 
-# if lintrunner is not installed, install it
-if ! command -v lintrunner &> /dev/null; then
-    python3 -m pip install lintrunner==0.12.5
-fi
-
 # This has already been cached in the docker image
 lintrunner init 2> /dev/null
 
@@ -38,11 +33,10 @@ python3 torch/utils/data/datapipes/gen_pyi.py
 
 RC=0
 # Run lintrunner on all files
-if ! lintrunner --force-color --tee-json=lint.json ${ADDITIONAL_LINTRUNNER_ARGS} 2> /dev/null; then
+if ! lintrunner --force-color --all-files --tee-json=lint.json ${ADDITIONAL_LINTRUNNER_ARGS} 2> /dev/null; then
     echo ""
     echo -e "\e[1m\e[36mYou can reproduce these results locally by using \`lintrunner -m origin/main\`. (If you don't get the same results, run \'lintrunner init\' to update your local linter)\e[0m"
-    echo -e "\e[1m\e[36mSee https://github.com/pytorch/pytorch/wiki/lintrunner for setup instructions. To apply suggested patches automatically, use the -a flag. Before pushing another commit,\e[0m"
-    echo -e "\e[1m\e[36mplease verify locally and ensure everything passes.\e[0m"
+    echo -e "\e[1m\e[36mSee https://github.com/pytorch/pytorch/wiki/lintrunner for setup instructions.\e[0m"
     RC=1
 fi
 

.github/scripts/runner_determinator.py

@@ -1,9 +1,5 @@
 # flake8: noqa: G004
 
-# Note: Copies of this script in runner_determinator.py and _runner-determinator.yml
-#       must be kept in sync. You can do it easily by running the following command:
-#           python .github/scripts/update_runner_determinator.py
-
 """
 This runner determinator is used to determine which set of runners to run a
 GitHub job on. It uses the first comment of a GitHub issue (by default
@@ -39,8 +35,7 @@ Example config:
     experiments:
       lf:
         rollout_percent: 25
-        all_branches: false
-        default: true
+
     ---
 
     # Opt-ins:
@@ -58,7 +53,7 @@ import os
 import random
 from argparse import ArgumentParser
 from logging import LogRecord
-from typing import Any, Dict, FrozenSet, Iterable, List, NamedTuple, Tuple
+from typing import Any, Dict, Iterable, List, NamedTuple, Tuple
 
 import yaml
 from github import Auth, Github
@@ -84,12 +79,6 @@ class Experiment(NamedTuple):
     rollout_perc: float = (
         0  # Percentage of workflows to experiment on when user is not opted-in.
     )
-    all_branches: bool = (
-        False  # If True, the experiment is also enabled on the exception branches
-    )
-    default: bool = (
-        True  # If True, the experiment is enabled by default for all queries
-    )
 
     # Add more fields as needed
 
@@ -144,12 +133,6 @@ def set_github_output(key: str, value: str) -> None:
         f.write(f"{key}={value}\n")
 
 
-def _str_comma_separated_to_set(value: str) -> FrozenSet[str]:
-    return frozenset(
-        filter(lambda itm: itm != "", map(str.strip, value.strip(" \n\t").split(",")))
-    )
-
-
 def parse_args() -> Any:
     parser = ArgumentParser("Get dynamic rollout settings")
     parser.add_argument("--github-token", type=str, required=True, help="GitHub token")
@@ -184,13 +167,6 @@ def parse_args() -> Any:
         required=True,
         help="Current GitHub ref type, branch or tag",
     )
-    parser.add_argument(
-        "--eligible-experiments",
-        type=_str_comma_separated_to_set,
-        required=False,
-        default="",
-        help="comma separated list of experiments to check, if omitted all experiments marked with default=True are checked",
-    )
 
     return parser.parse_args()
 
@@ -236,7 +212,7 @@ def get_potential_pr_author(
 
 def is_exception_branch(branch: str) -> bool:
     """
-    Branches that get opted out of experiments by default, until they're explicitly enabled.
+    Branches that get opted out of all experiments and should always use Meta runners
     """
     return branch.split("/")[0] in {"main", "nightly", "release", "landchecks"}
 
@@ -362,11 +338,7 @@ def is_user_opted_in(user: str, user_optins: UserOptins, experiment_name: str) -
 
 
 def get_runner_prefix(
-    rollout_state: str,
-    workflow_requestors: Iterable[str],
-    branch: str,
-    eligible_experiments: FrozenSet[str] = frozenset(),
-    is_canary: bool = False,
+    rollout_state: str, workflow_requestors: Iterable[str], is_canary: bool = False
 ) -> str:
     settings = parse_settings(rollout_state)
     user_optins = parse_users(rollout_state)
@@ -374,24 +346,7 @@ def get_runner_prefix(
     fleet_prefix = ""
     prefixes = []
     for experiment_name, experiment_settings in settings.experiments.items():
-        if not experiment_settings.all_branches and is_exception_branch(branch):
-            log.info(
-                f"Branch {branch} is an exception branch. Not enabling experiment {experiment_name}."
-            )
-            continue
-
-        if eligible_experiments:
-            if experiment_name not in eligible_experiments:
-                exp_list = ", ".join(eligible_experiments)
-                log.info(
-                    f"Skipping experiment '{experiment_name}', as it is not in the eligible_experiments list: {exp_list}"
-                )
-                continue
-        elif not experiment_settings.default:
-            log.info(
-                f"Skipping experiment '{experiment_name}', as it is not a default experiment"
-            )
-            continue
+        enabled = False
 
         # Is any workflow_requestor opted in to this experiment?
         opted_in_users = [
@@ -400,13 +355,11 @@ def get_runner_prefix(
             if is_user_opted_in(requestor, user_optins, experiment_name)
         ]
 
-        enabled = False
         if opted_in_users:
             log.info(
                 f"{', '.join(opted_in_users)} have opted into experiment {experiment_name}."
             )
             enabled = True
-
         elif experiment_settings.rollout_perc:
             # If no user is opted in, then we randomly enable the experiment based on the rollout percentage
             if random.uniform(0, 100) <= experiment_settings.rollout_perc:
@@ -454,35 +407,35 @@ def get_rollout_state_from_issue(github_token: str, repo: str, issue_num: int) -
 def main() -> None:
     args = parse_args()
 
-    runner_label_prefix = DEFAULT_LABEL_PREFIX
-
-    try:
-        rollout_state = get_rollout_state_from_issue(
-            args.github_token, args.github_issue_repo, args.github_issue
+    if args.github_ref_type == "branch" and is_exception_branch(args.github_branch):
+        log.info(
+            f"Exception branch: '{args.github_branch}', using Meta runners and no experiments."
         )
+        runner_label_prefix = DEFAULT_LABEL_PREFIX
+    else:
+        try:
+            rollout_state = get_rollout_state_from_issue(
+                args.github_token, args.github_issue_repo, args.github_issue
+            )
 
-        username = get_potential_pr_author(
-            args.github_token,
-            args.github_repo,
-            args.github_actor,
-            args.github_ref_type,
-            args.github_branch,
-        )
+            username = get_potential_pr_author(
+                args.github_token,
+                args.github_repo,
+                args.github_actor,
+                args.github_ref_type,
+                args.github_branch,
+            )
 
-        is_canary = args.github_repo == "pytorch/pytorch-canary"
+            is_canary = args.github_repo == "pytorch/pytorch-canary"
 
-        runner_label_prefix = get_runner_prefix(
-            rollout_state,
-            (args.github_issue_owner, username),
-            args.github_branch,
-            args.eligible_experiments,
-            is_canary,
-        )
+            runner_label_prefix = get_runner_prefix(
+                rollout_state, (args.github_issue_owner, username), is_canary
+            )
 
-    except Exception as e:
-        log.error(
-            f"Failed to get issue. Defaulting to Meta runners and no experiments. Exception: {e}"
-        )
+        except Exception as e:
+            log.error(
+                f"Failed to get issue. Defaulting to Meta runners and no experiments. Exception: {e}"
+            )
 
     set_github_output(GH_OUTPUT_KEY_LABEL_TYPE, runner_label_prefix)
 

.github/scripts/sync_distributed_folder_prototype.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+set -eoux pipefail
+
+SYNC_BRANCH=pytorch-stable-prototype
+
+git config user.email "fake@example.com"
+git config user.name  "PyTorch Stable Bot"
+
+git fetch origin main
+git fetch origin "$SYNC_BRANCH"
+git checkout "$SYNC_BRANCH"
+
+# Using a hardcoded SHA here is a massive speedup as we can skip the entire history of the pytorch GitHub repo.
+# This specific SHA was chosen as it was before the "branch point" of the stable branch
+for SHA in $(git log ba3b05fdf37ddbc3c301294d6a560a816335e717..origin/main --pretty="%h" -- torch/distributed torch/csrc/distributed test/distributed test/cpp/c10d benchmarks/distributed)
+do
+    # `git merge-base --is-ancestor` exits with code 0 if the given SHA is an ancestor, and non-0 otherwise
+    if git merge-base --is-ancestor $SHA HEAD || [[ $(git log --grep="(cherry picked from commit $SHA") ]]
+    then
+        echo "Skipping $SHA"
+        continue
+    fi
+    echo "Copying $SHA"
+    git cherry-pick -x "$SHA" -X theirs
+    git reset --soft HEAD~1
+    git add torch/distributed torch/csrc/distributed test/distributed test/cpp/c10d benchmarks/distributed
+    git checkout .
+    git commit --reuse-message=HEAD@{1}
+    git clean -f
+done
+
+if [[ "${WITH_PUSH}" == true ]]; then
+  git push
+fi

.github/scripts/test_runner_determinator.py

@@ -4,10 +4,6 @@ from unittest.mock import Mock, patch
 import runner_determinator as rd
 
 
-USER_BRANCH = "somebranch"
-EXCEPTION_BRANCH = "main"
-
-
 class TestRunnerDeterminatorIssueParser(TestCase):
     def test_parse_settings(self) -> None:
         settings_text = """
@@ -16,7 +12,6 @@ class TestRunnerDeterminatorIssueParser(TestCase):
                 rollout_perc: 25
             otherExp:
                 rollout_perc: 0
-                default: false
         ---
 
         Users:
@@ -33,7 +28,7 @@ class TestRunnerDeterminatorIssueParser(TestCase):
             "lf settings not parsed correctly",
         )
         self.assertTupleEqual(
-            rd.Experiment(rollout_perc=0, default=False),
+            rd.Experiment(rollout_perc=0),
             settings.experiments["otherExp"],
             "otherExp settings not parsed correctly",
         )
@@ -47,7 +42,7 @@ class TestRunnerDeterminatorIssueParser(TestCase):
                 rollout_perc: 25
             otherExp:
                 rollout_perc: 0
-                default: false
+
         ```
 
         ---
@@ -66,41 +61,7 @@ class TestRunnerDeterminatorIssueParser(TestCase):
             "lf settings not parsed correctly",
         )
         self.assertTupleEqual(
-            rd.Experiment(rollout_perc=0, default=False),
-            settings.experiments["otherExp"],
-            "otherExp settings not parsed correctly",
-        )
-
-    def test_parse_all_branches_setting(self) -> None:
-        settings_text = """
-        ```
-        experiments:
-            lf:
-                rollout_perc: 25
-                all_branches: true
-            otherExp:
-                all_branches: True
-                rollout_perc: 0
-        ```
-
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-
-        settings = rd.parse_settings(settings_text)
-
-        self.assertTupleEqual(
-            rd.Experiment(rollout_perc=25, all_branches=True),
-            settings.experiments["lf"],
-            "lf settings not parsed correctly",
-        )
-        self.assertTrue(settings.experiments["otherExp"].all_branches)
-        self.assertTupleEqual(
-            rd.Experiment(rollout_perc=0, all_branches=True),
+            rd.Experiment(rollout_perc=0),
             settings.experiments["otherExp"],
             "otherExp settings not parsed correctly",
         )
@@ -158,7 +119,7 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
         @User2,lf,otherExp
 
         """
-        prefix = rd.get_runner_prefix(settings_text, ["User1"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User1"])
         self.assertEqual("lf.", prefix, "Runner prefix not correct for User1")
 
     def test_opted_in_user_two_experiments(self) -> None:
@@ -175,67 +136,9 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
         @User2,lf,otherExp
 
         """
-        prefix = rd.get_runner_prefix(settings_text, ["User2"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User2"])
         self.assertEqual("lf.otherExp.", prefix, "Runner prefix not correct for User2")
 
-    def test_opted_in_user_two_experiments_default(self) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 0
-            otherExp:
-                rollout_perc: 0
-                default: false
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-        prefix = rd.get_runner_prefix(settings_text, ["User2"], USER_BRANCH)
-        self.assertEqual("lf.", prefix, "Runner prefix not correct for User2")
-
-    def test_opted_in_user_two_experiments_default_exp(self) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 0
-            otherExp:
-                rollout_perc: 0
-                default: false
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-        prefix = rd.get_runner_prefix(
-            settings_text, ["User2"], USER_BRANCH, frozenset(["lf", "otherExp"])
-        )
-        self.assertEqual("lf.otherExp.", prefix, "Runner prefix not correct for User2")
-
-    def test_opted_in_user_two_experiments_default_exp_2(self) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 0
-            otherExp:
-                rollout_perc: 0
-                default: false
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-        prefix = rd.get_runner_prefix(
-            settings_text, ["User2"], USER_BRANCH, frozenset(["otherExp"])
-        )
-        self.assertEqual("otherExp.", prefix, "Runner prefix not correct for User2")
-
     @patch("random.uniform", return_value=50)
     def test_opted_out_user(self, mock_uniform: Mock) -> None:
         settings_text = """
@@ -251,7 +154,7 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
         @User2,lf,otherExp
 
         """
-        prefix = rd.get_runner_prefix(settings_text, ["User3"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User3"])
         self.assertEqual("", prefix, "Runner prefix not correct for user")
 
     @patch("random.uniform", return_value=10)
@@ -271,80 +174,9 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
         """
 
         # User3 is opted out, but is pulled into both experiments by the 10% rollout
-        prefix = rd.get_runner_prefix(settings_text, ["User3"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User3"])
         self.assertEqual("lf.otherExp.", prefix, "Runner prefix not correct for user")
 
-    @patch("random.uniform", return_value=10)
-    def test_opted_out_user_was_pulled_in_by_rollout_excl_nondefault(
-        self, mock_uniform: Mock
-    ) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 25
-            otherExp:
-                rollout_perc: 25
-                default: false
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-
-        # User3 is opted out, but is pulled into default experiments by the 10% rollout
-        prefix = rd.get_runner_prefix(settings_text, ["User3"], USER_BRANCH)
-        self.assertEqual("lf.", prefix, "Runner prefix not correct for user")
-
-    @patch("random.uniform", return_value=10)
-    def test_opted_out_user_was_pulled_in_by_rollout_filter_exp(
-        self, mock_uniform: Mock
-    ) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 25
-            otherExp:
-                rollout_perc: 25
-                default: false
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-
-        # User3 is opted out, but is pulled into default experiments by the 10% rollout
-        prefix = rd.get_runner_prefix(
-            settings_text, ["User3"], USER_BRANCH, frozenset(["otherExp"])
-        )
-        self.assertEqual("otherExp.", prefix, "Runner prefix not correct for user")
-
-    @patch("random.uniform", return_value=25)
-    def test_opted_out_user_was_pulled_out_by_rollout_filter_exp(
-        self, mock_uniform: Mock
-    ) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 10
-            otherExp:
-                rollout_perc: 50
-                default: false
-        ---
-
-        Users:
-        @User1,lf
-        @User2,lf,otherExp
-
-        """
-
-        # User3 is opted out, but is pulled into default experiments by the 10% rollout
-        prefix = rd.get_runner_prefix(settings_text, ["User3"], USER_BRANCH)
-        self.assertEqual("", prefix, "Runner prefix not correct for user")
-
     def test_lf_prefix_always_comes_first(self) -> None:
         settings_text = """
         experiments:
@@ -360,7 +192,7 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
 
         """
 
-        prefix = rd.get_runner_prefix(settings_text, ["User2"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User2"])
         self.assertEqual("lf.otherExp.", prefix, "Runner prefix not correct for user")
 
     def test_ignores_commented_users(self) -> None:
@@ -378,7 +210,7 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
 
         """
 
-        prefix = rd.get_runner_prefix(settings_text, ["User1"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User1"])
         self.assertEqual("", prefix, "Runner prefix not correct for user")
 
     def test_ignores_extra_experiments(self) -> None:
@@ -397,44 +229,9 @@ class TestRunnerDeterminatorGetRunnerPrefix(TestCase):
 
         """
 
-        prefix = rd.get_runner_prefix(settings_text, ["User1"], USER_BRANCH)
+        prefix = rd.get_runner_prefix(settings_text, ["User1"])
         self.assertEqual("lf.otherExp.", prefix, "Runner prefix not correct for user")
 
-    def test_disables_experiment_on_exception_branches_when_not_explicitly_opted_in(
-        self,
-    ) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 100
-        ---
-
-        Users:
-        @User,lf,otherExp
-
-        """
-
-        prefix = rd.get_runner_prefix(settings_text, ["User1"], EXCEPTION_BRANCH)
-        self.assertEqual("", prefix, "Runner prefix not correct for user")
-
-    def test_allows_experiment_on_exception_branches_when_explicitly_opted_in(
-        self,
-    ) -> None:
-        settings_text = """
-        experiments:
-            lf:
-                rollout_perc: 100
-                all_branches: true
-        ---
-
-        Users:
-        @User,lf,otherExp
-
-        """
-
-        prefix = rd.get_runner_prefix(settings_text, ["User1"], EXCEPTION_BRANCH)
-        self.assertEqual("lf.", prefix, "Runner prefix not correct for user")
-
 
 if __name__ == "__main__":
     main()

.github/scripts/test_trymerge.py

@@ -12,7 +12,7 @@ import json
 import os
 import warnings
 from hashlib import sha256
-from typing import Any, List, Optional
+from typing import Any, Dict, List, Optional
 from unittest import main, mock, skip, TestCase
 from urllib.error import HTTPError
 
@@ -24,6 +24,7 @@ from trymerge import (
     find_matching_merge_rule,
     get_classifications,
     get_drci_classifications,
+    get_rockset_results,
     gh_get_team_members,
     GitHubPR,
     JobCheckState,
@@ -41,6 +42,7 @@ if "GIT_REMOTE_URL" not in os.environ:
     os.environ["GIT_REMOTE_URL"] = "https://github.com/pytorch/pytorch"
 
 GQL_MOCKS = "gql_mocks.json.gz"
+ROCKSET_MOCKS = "rockset_mocks.json.gz"
 DRCI_MOCKS = "drci_mocks.json.gz"
 
 
@@ -75,11 +77,16 @@ def mock_query(
         if err.code == 401 or err.code == 403:
             err_msg = f"If you are seeing this message during workflow run, please make sure to update {file_name}"
             err_msg += f" locally, by deleting it and running {os.path.basename(__file__)} with"
-            err_msg += " GitHub Personal Access Token passed via GITHUB_TOKEN"
+            err_msg += " GitHub Personal Access Token passed via GITHUB_TOKEN,"
+            err_msg += " the rockset api key passed via ROCKSET_API_KEY,"
             err_msg += " and drci api key passed via DRCI_BOT_KEY environment variables"
-            if os.getenv("GITHUB_TOKEN") is None or os.getenv("DRCI_BOT_KEY") is None:
+            if (
+                os.getenv("GITHUB_TOKEN") is None
+                or os.getenv("ROCKSET_API_KEY") is None
+                or os.getenv("DRCI_BOT_KEY") is None
+            ):
                 err_msg = (
-                    "Failed to update cached queries as GITHUB_TOKEN or DRCI_BOT_KEY "
+                    "Failed to update cached queries as GITHUB_TOKEN or ROCKSET_API_KEY or DRCI_BOT_KEY "
                     + "is not defined. "
                     + err_msg
                 )
@@ -103,6 +110,16 @@ def mocked_gh_graphql(query: str, **kwargs: Any) -> Any:
     return mock_query(gh_graphql_wrapper, GQL_MOCKS, key_function, query, kwargs)
 
 
+def mocked_rockset_results(head_sha: str, merge_base: str, num_retries: int = 3) -> Any:
+    return mock_query(
+        get_rockset_results,
+        ROCKSET_MOCKS,
+        lambda x, y: f"{x} {y}",
+        head_sha,
+        merge_base,
+    )
+
+
 def mocked_drci_classifications(pr_num: int, project: str, num_retries: int = 3) -> Any:
     return mock_query(
         get_drci_classifications,
@@ -256,6 +273,10 @@ def xla_merge_rules(repo: Any, org: str, project: str) -> List[MergeRule]:
     ]
 
 
+def empty_rockset_results(head_sha: str, merge_base: str) -> List[Dict[str, Any]]:
+    return []
+
+
 class DummyGitRepo(GitRepo):
     def __init__(self) -> None:
         super().__init__(get_git_repo_dir(), get_git_remote_name())
@@ -267,6 +288,7 @@ class DummyGitRepo(GitRepo):
         return "super awsome commit message"
 
 
+@mock.patch("trymerge.get_rockset_results", side_effect=empty_rockset_results)
 @mock.patch("trymerge.gh_graphql", side_effect=mocked_gh_graphql)
 @mock.patch(
     "trymerge.get_drci_classifications", side_effect=mocked_drci_classifications
@@ -582,6 +604,7 @@ class TestTryMerge(TestCase):
             mocked_gh_fetch_merge_base.assert_called_once()
 
 
+@mock.patch("trymerge.get_rockset_results", side_effect=mocked_rockset_results)
 @mock.patch("trymerge.gh_graphql", side_effect=mocked_gh_graphql)
 @mock.patch("trymerge.gh_fetch_merge_base", return_value="")
 @mock.patch(
@@ -820,7 +843,7 @@ class TestBypassFailures(TestCase):
         checks = pr.get_checkrun_conclusions()
 
         # Known flaky failure takes precedence over ignore current (need to set the
-        # merge base here to get the results from Dr. CI, and that categorize the
+        # merge base here to get the results from Rockset, and that categorize the
         # broken trunk failure too
         checks = get_classifications(
             pr.pr_num,
@@ -906,6 +929,7 @@ class TestBypassFailures(TestCase):
         )
 
 
+@mock.patch("trymerge.get_rockset_results", side_effect=mocked_rockset_results)
 @mock.patch("trymerge.gh_graphql", side_effect=mocked_gh_graphql)
 @mock.patch("trymerge.gh_fetch_merge_base", return_value="")
 @mock.patch("trymerge.get_drci_classifications", return_value={})
@@ -984,6 +1008,7 @@ class TestBypassFailuresOnSandCastle(TestCase):
         self.assertTrue(len(failed) == 2)
 
 
+@mock.patch("trymerge.get_rockset_results", side_effect=mocked_rockset_results)
 @mock.patch("trymerge.gh_graphql", side_effect=mocked_gh_graphql)
 @mock.patch("trymerge.gh_fetch_merge_base", return_value="")
 @mock.patch(

.github/scripts/trymerge.py

@@ -452,6 +452,8 @@ RE_DIFF_REV = re.compile(r"^Differential Revision:.+?(D[0-9]+)", re.MULTILINE)
 CIFLOW_LABEL = re.compile(r"^ciflow/.+")
 CIFLOW_TRUNK_LABEL = re.compile(r"^ciflow/trunk")
 MERGE_RULE_PATH = Path(".github") / "merge_rules.yaml"
+ROCKSET_MERGES_COLLECTION = "merges"
+ROCKSET_MERGES_WORKSPACE = "commons"
 REMOTE_MAIN_BRANCH = "origin/main"
 DRCI_CHECKRUN_NAME = "Dr.CI"
 INTERNAL_CHANGES_CHECKRUN_NAME = "Meta Internal-Only Changes Check"
@@ -1178,7 +1180,7 @@ class GitHubPR:
         merge_commit_sha = repo.rev_parse(name=self.default_branch())
 
         if comment_id and self.pr_num:
-            # Finally, upload the record to s3. The list of pending and failed
+            # Finally, upload the record to Rockset. The list of pending and failed
             # checks are at the time of the merge
             save_merge_record(
                 comment_id=comment_id,
@@ -1200,7 +1202,7 @@ class GitHubPR:
                 ignore_current=bool(ignore_current_checks),
             )
         else:
-            print("Missing comment ID or PR number, couldn't upload to s3")
+            print("Missing comment ID or PR number, couldn't upload to Rockset")
 
         # Usually Github will see that the commit has "resolves <pr_num>" in the
         # commit message and close the PR, but sometimes it doesn't, leading to
@@ -1479,7 +1481,7 @@ def find_matching_merge_rule(
 
         # Categorize all checks when skip_mandatory_checks (force merge) is set. Do it here
         # where the list of checks is readily available. These records will be saved into
-        # s3 merge records
+        # Rockset merge records
         (
             pending_mandatory_checks,
             failed_mandatory_checks,
@@ -1506,7 +1508,7 @@ def checks_to_str(checks: List[Tuple[str, Optional[str]]]) -> str:
 
 
 def checks_to_markdown_bullets(
-    checks: List[Tuple[str, Optional[str], Optional[int]]],
+    checks: List[Tuple[str, Optional[str], Optional[int]]]
 ) -> List[str]:
     return [
         f"- [{c[0]}]({c[1]})" if c[1] is not None else f"- {c[0]}" for c in checks[:5]
@@ -1566,7 +1568,7 @@ def save_merge_record(
     This saves the merge records as a json, which can later be uploaded to s3
     """
 
-    # Prepare the record to be written into s3
+    # Prepare the record to be written into Rockset
     data = [
         {
             "comment_id": comment_id,
@@ -1588,8 +1590,7 @@ def save_merge_record(
             "ignore_current": ignore_current,
             "error": error,
             # This is a unique identifier for the record for deduping purposes
-            # in Rockset.  Any unique string would work.  This will not be used
-            # after we migrate off Rockset
+            # in rockset.  Any unique string would work
             "_id": f"{project}-{pr_num}-{comment_id}-{os.environ.get('GITHUB_RUN_ID')}",
         }
     ]
@@ -1599,6 +1600,36 @@ def save_merge_record(
         json.dump(data, f)
 
 
+@retries_decorator(rc=[])
+def get_rockset_results(head_sha: str, merge_base: str) -> List[Dict[str, Any]]:
+    query = f"""
+SELECT
+    w.name as workflow_name,
+    j.id,
+    j.name,
+    j.conclusion,
+    j.completed_at,
+    j.html_url,
+    j.head_sha,
+    j.torchci_classification.captures as failure_captures,
+    LENGTH(j.steps) as steps,
+FROM
+    commons.workflow_job j join commons.workflow_run w on w.id = j.run_id
+where
+    j.head_sha in ('{head_sha}','{merge_base}')
+"""
+    try:
+        import rockset  # type: ignore[import]
+
+        res = rockset.RocksetClient(
+            host="api.usw2a1.rockset.com", api_key=os.environ["ROCKSET_API_KEY"]
+        ).sql(query)
+        return cast(List[Dict[str, Any]], res.results)
+    except ModuleNotFoundError:
+        print("Could not use RockSet as rocket dependency is missing")
+        return []
+
+
 @retries_decorator()
 def get_drci_classifications(pr_num: int, project: str = "pytorch") -> Any:
     """
@@ -2036,7 +2067,7 @@ def categorize_checks(
     pending_checks: List[Tuple[str, Optional[str], Optional[int]]] = []
     failed_checks: List[Tuple[str, Optional[str], Optional[int]]] = []
 
-    # failed_checks_categorization is used to keep track of all ignorable failures when saving the merge record on s3
+    # failed_checks_categorization is used to keep track of all ignorable failures when saving the merge record on Rockset
     failed_checks_categorization: Dict[str, List[Any]] = defaultdict(list)
 
     # If required_checks is not set or empty, consider all names are relevant
@@ -2095,7 +2126,7 @@ def categorize_checks(
     ):
         failed_checks = failed_checks + flaky_or_broken_trunk
 
-    # The list of failed_checks_categorization is returned so that it can be saved into the s3 merge record
+    # The list of failed_checks_categorization is returned so that it can be saved into the Rockset merge record
     return (pending_checks, failed_checks, failed_checks_categorization)
 
 
@@ -2379,7 +2410,7 @@ def main() -> None:
         handle_exception(e)
 
         if args.comment_id and args.pr_num:
-            # Finally, upload the record to s3, we don't have access to the
+            # Finally, upload the record to Rockset, we don't have access to the
             # list of pending and failed checks here, but they are not really
             # needed at the moment
             save_merge_record(
@@ -2402,7 +2433,7 @@ def main() -> None:
                 error=str(e),
             )
         else:
-            print("Missing comment ID or PR number, couldn't upload to s3")
+            print("Missing comment ID or PR number, couldn't upload to Rockset")
     finally:
         if not args.check_mergeability:
             gh_remove_label(

.github/scripts/update_runner_determinator.py

@@ -1,31 +0,0 @@
-#!/usr/bin/env python3
-
-import re
-
-
-# Read the contents of runner_determinator.py
-with open(".github/scripts/runner_determinator.py") as script_file:
-    script_content = script_file.read()
-
-# Indent the script content by 10 spaces to match destination indentation
-indented_script_content = "\n".join(
-    [" " * 10 + line if line else line for line in script_content.splitlines()]
-)
-
-# Read the contents of _runner-determinator.yml
-with open(".github/workflows/_runner-determinator.yml") as yml_file:
-    yml_content = yml_file.read()
-
-# Replace the content between the markers
-new_yml_content = re.sub(
-    r"(cat <<EOF > runner_determinator.py\n)(.*?)(\n\s+EOF)",
-    lambda match: match.group(1) + indented_script_content + match.group(3),
-    yml_content,
-    flags=re.DOTALL,
-)
-
-# Save the modified content back to _runner-determinator.yml
-with open(".github/workflows/_runner-determinator.yml", "w") as yml_file:
-    yml_file.write(new_yml_content)
-
-print("Updated _runner-determinator.yml with the contents of runner_determinator.py")

.github/templates/linux_binary_build_workflow.yml.j2

@@ -54,7 +54,7 @@ env:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -68,7 +68,6 @@ jobs:
     needs: get-label-type
     with:!{{ upload.binary_env_as_input(config) }}
       {%- if "aarch64" in build_environment %}
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       {%- elif "s390x" in build_environment %}
@@ -103,7 +102,6 @@ jobs:
       build_name: !{{ config["build_name"] }}
       build_environment: !{{ build_environment }}
       {%- if "aarch64" in build_environment %}
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.2xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
       {%- elif "s390x" in build_environment %}

.github/templates/windows_binary_build_workflow.yml.j2

@@ -55,7 +55,7 @@ env:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/_bazel-build-test.yml

@@ -91,14 +91,14 @@ jobs:
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
-      - name: Check if in a container runner
+      - name: Check if in a ARC runner
         shell: bash
-        id: check_container_runner
-        run: echo "IN_CONTAINER_RUNNER=$(if [ -f /.inarc ] || [ -f /.incontainer ]; then echo true ; else echo false; fi)" >> "$GITHUB_OUTPUT"
+        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
         uses: pytorch/test-infra/.github/actions/setup-nvidia@main
-        if: ${{ inputs.cuda-version != 'cpu' && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
+        if: ${{ inputs.cuda-version != 'cpu' && steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
 
       - name: Output disk space left
         run: |

.github/workflows/_binary-build-linux.yml

@@ -272,8 +272,6 @@ jobs:
           docker exec -t -w "${PYTORCH_ROOT}" "${container_name}" bash -c "bash .circleci/scripts/binary_populate_env.sh"
           if [[ ${BUILD_ENVIRONMENT} == *"aarch64"* ]]; then
             docker exec -t "${container_name}" bash -c "bash /builder/aarch64_linux/aarch64_ci_build.sh"
-          elif [[ ${{ inputs.PACKAGE_TYPE }} == "manywheel" || ${{ inputs.PACKAGE_TYPE }} == "libtorch" ]]; then
-            docker exec -t "${container_name}" bash -c "source ${BINARY_ENV_FILE} && bash /pytorch/.ci/${{ inputs.PACKAGE_TYPE }}/build.sh"
           else
             docker exec -t "${container_name}" bash -c "source ${BINARY_ENV_FILE} && bash /builder/${{ inputs.PACKAGE_TYPE }}/build.sh"
           fi

.github/workflows/_linux-build.yml

@@ -109,7 +109,6 @@ jobs:
     steps:
       - name: Setup SSH (Click me for login details)
         uses: pytorch/test-infra/.github/actions/setup-ssh@main
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
@@ -119,16 +118,13 @@ jobs:
       # checkout. In other cases you should prefer a local checkout.
       - name: Checkout PyTorch
         uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
-        with:
-          no-sudo: ${{ inputs.build-environment == 'linux-s390x-binary-manywheel' }}
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
 
       - name: configure aws credentials
         uses: aws-actions/configure-aws-credentials@v3
-        if: ${{ inputs.aws-role-to-assume != '' && inputs.build-environment != 'linux-s390x-binary-manywheel' }}
+        if: ${{ inputs.aws-role-to-assume != '' }}
         with:
           role-to-assume: ${{ inputs.aws-role-to-assume }}
           role-session-name: gha-linux-build
@@ -137,13 +133,11 @@ jobs:
       - name: Calculate docker image
         id: calculate-docker-image
         uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           docker-image-name: ${{ inputs.docker-image-name }}
 
       - name: Use following to pull public copy of the image
         id: print-ghcr-mirror
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         env:
           ECR_DOCKER_IMAGE: ${{ steps.calculate-docker-image.outputs.docker-image }}
         shell: bash
@@ -153,7 +147,6 @@ jobs:
 
       - name: Pull docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -181,7 +174,6 @@ jobs:
       - name: Download pytest cache
         uses: ./.github/actions/pytest-cache-download
         continue-on-error: true
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           cache_dir: .pytest_cache
           job_identifier: ${{ github.workflow }}_${{ inputs.build-environment }}
@@ -203,7 +195,6 @@ jobs:
           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 }}
-          DOCKER_IMAGE_S390X: ${{ inputs.docker-image-name }}
           XLA_CUDA: ${{ contains(inputs.build-environment, 'xla') && '0' || '' }}
           DEBUG: ${{ inputs.build-with-debug && '1' || '0' }}
           OUR_GITHUB_JOB_ID: ${{ steps.get-job-id.outputs.job-id }}
@@ -211,21 +202,7 @@ jobs:
           SCRIBE_GRAPHQL_ACCESS_TOKEN: ${{ secrets.SCRIBE_GRAPHQL_ACCESS_TOKEN }}
           USE_SPLIT_BUILD: ${{ inputs.use_split_build }}
         run: |
-          if [[ ${BUILD_ENVIRONMENT} == *"s390x"* ]]; then
-            JENKINS_USER=
-            USED_IMAGE="${DOCKER_IMAGE_S390X}"
-
-            # since some steps are skipped on s390x, if they are necessary, run them here
-            env | grep '^GITHUB' >> "/tmp/github_env_${GITHUB_RUN_ID}"
-            env | grep '^CI' >> "/tmp/github_env_${GITHUB_RUN_ID}"
-          else
-            JENKINS_USER="--user jenkins"
-            USED_IMAGE="${DOCKER_IMAGE}"
-          fi
-
           # detached container should get cleaned up by teardown_ec2_linux
-          # Used for JENKINS_USER, which can be empty
-          # shellcheck disable=SC2086
           container_name=$(docker run \
             -e BUILD_ENVIRONMENT \
             -e MAX_JOBS="$(nproc --ignore=2)" \
@@ -248,10 +225,10 @@ jobs:
             --cap-add=SYS_PTRACE \
             --tty \
             --detach \
-            ${JENKINS_USER} \
+            --user jenkins \
             -v "${GITHUB_WORKSPACE}:/var/lib/jenkins/workspace" \
             -w /var/lib/jenkins/workspace \
-            "${USED_IMAGE}"
+            "${DOCKER_IMAGE}"
           )
           docker exec -t "${container_name}" sh -c '.ci/pytorch/build.sh'
 
@@ -262,7 +239,7 @@ jobs:
 
       - name: Store PyTorch Build Artifacts on S3
         uses: seemethere/upload-artifact-s3@v5
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && !inputs.use_split_build && inputs.build-environment != 'linux-s390x-binary-manywheel'
+        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && !inputs.use_split_build
         with:
           name: ${{ inputs.build-environment }}
           retention-days: 14
@@ -272,7 +249,7 @@ jobs:
 
       - name: Store PyTorch Build Artifacts on S3 for split build
         uses: seemethere/upload-artifact-s3@v5
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.use_split_build && inputs.build-environment != 'linux-s390x-binary-manywheel'
+        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.use_split_build
         with:
           name: ${{ inputs.build-environment }}-experimental-split-build
           retention-days: 14
@@ -280,26 +257,8 @@ jobs:
           path: artifacts.zip
           s3-bucket: ${{ inputs.s3-bucket }}
 
-      - name: Store PyTorch Build Artifacts for s390x
-        uses: actions/upload-artifact@v3
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && !inputs.use_split_build && inputs.build-environment == 'linux-s390x-binary-manywheel'
-        with:
-          name: ${{ inputs.build-environment }}
-          retention-days: 14
-          if-no-files-found: error
-          path: artifacts.zip
-
-      - name: Store PyTorch Build Artifacts for s390x for split build
-        uses: actions/upload-artifact@v3
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.use_split_build && inputs.build-environment == 'linux-s390x-binary-manywheel'
-        with:
-          name: ${{ inputs.build-environment }}-experimental-split-build
-          retention-days: 14
-          if-no-files-found: error
-          path: artifacts.zip
-
       - name: Upload sccache stats
-        if: steps.build.outcome != 'skipped' && inputs.build-environment != 'linux-s390x-binary-manywheel'
+        if: steps.build.outcome != 'skipped'
         uses: seemethere/upload-artifact-s3@v5
         with:
           s3-prefix: |
@@ -311,13 +270,4 @@ jobs:
 
       - name: Teardown Linux
         uses: pytorch/test-infra/.github/actions/teardown-linux@main
-        if: always() && inputs.build-environment != 'linux-s390x-binary-manywheel'
-
-      - name: Cleanup docker
-        if: always() && inputs.build-environment == 'linux-s390x-binary-manywheel'
-        shell: bash
-        run: |
-          # on s390x stop the container for clean worker stop
-          # ignore expansion of "docker ps -q" since it could be empty
-          # shellcheck disable=SC2046
-          docker stop $(docker ps -q) || true
+        if: always()

.github/workflows/_linux-test.yml

@@ -114,32 +114,22 @@ jobs:
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
-      - name: Check if in a container runner
+      - name: Check if in a ARC runner
         shell: bash
-        id: check_container_runner
-        run: echo "IN_CONTAINER_RUNNER=$(if [ -f /.inarc ] || [ -f /.incontainer ]; then echo true ; else echo false; fi)" >> "$GITHUB_OUTPUT"
+        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: ${{ contains(inputs.build-environment, 'cuda') && !contains(matrix.config, 'nogpu') && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
-
-      - name: Setup GPU_FLAG for docker run
-        id: setup-gpu-flag
-        run: echo "GPU_FLAG=--gpus all -e NVIDIA_DRIVER_CAPABILITIES=all" >> "${GITHUB_ENV}"
-        if: ${{ contains(inputs.build-environment, 'cuda') && !contains(matrix.config, 'nogpu') && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'true' }}
-
-      - name: Setup SCCACHE_SERVER_PORT environment for docker run when on container
-        id: setup-sscache-port-flag
-        run: echo "SCCACHE_SERVER_PORT_DOCKER_FLAG=-e SCCACHE_SERVER_PORT=$((RUNNER_UID + 4226))" >> "${GITHUB_ENV}"
-        if: ${{ steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'true' }}
+        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
         run: |
           sudo nvidia-smi -pm 1
           sudo nvidia-smi -ac 1215,1410
           nvidia-smi
-        if: ${{ contains(matrix.runner, 'a100') && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false' }}
+        if: contains(matrix.runner, 'a100')
 
       - name: Start monitoring script
         id: monitor-script
@@ -218,7 +208,6 @@ jobs:
           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_REGION: us-east-1
           SCCACHE_S3_KEY_PREFIX: ${{ github.workflow }}
           SHM_SIZE: ${{ contains(inputs.build-environment, 'cuda') && '2g' || '1g' }}
           DOCKER_IMAGE: ${{ inputs.docker-image }}
@@ -229,7 +218,6 @@ jobs:
           DASHBOARD_TAG: ${{ inputs.dashboard-tag }}
           HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
           SCRIBE_GRAPHQL_ACCESS_TOKEN: ${{ secrets.SCRIBE_GRAPHQL_ACCESS_TOKEN }}
-          IS_A100_RUNNER: ${{ contains(matrix.runner, 'a100') && '1' || '0' }}
 
         run: |
           set -x
@@ -248,7 +236,6 @@ jobs:
           # shellcheck disable=SC2086,SC2090
           container_name=$(docker run \
             ${GPU_FLAG:-} \
-            ${SCCACHE_SERVER_PORT_DOCKER_FLAG:-} \
             -e BUILD_ENVIRONMENT \
             -e PR_NUMBER \
             -e GITHUB_ACTIONS \
@@ -278,7 +265,6 @@ jobs:
             -e PR_LABELS \
             -e MAX_JOBS="$(nproc --ignore=2)" \
             -e SCCACHE_BUCKET \
-            -e SCCACHE_REGION \
             -e SCCACHE_S3_KEY_PREFIX \
             -e XLA_CUDA \
             -e XLA_CLANG_CACHE_S3_BUCKET_NAME \
@@ -288,7 +274,6 @@ jobs:
             -e HUGGING_FACE_HUB_TOKEN \
             -e SCRIBE_GRAPHQL_ACCESS_TOKEN \
             -e DASHBOARD_TAG \
-            -e IS_A100_RUNNER \
             --env-file="/tmp/github_env_${GITHUB_RUN_ID}" \
             --security-opt seccomp=unconfined \
             --cap-add=SYS_PTRACE \
@@ -358,7 +343,7 @@ jobs:
 
       - name: Teardown Linux
         uses: pytorch/test-infra/.github/actions/teardown-linux@main
-        if: always() && steps.check_container_runner.outputs.IN_CONTAINER_RUNNER == 'false'
+        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

.github/workflows/_mac-test-mps.yml

@@ -88,13 +88,6 @@ jobs:
           environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}
           pip-requirements-file: .github/requirements/pip-requirements-${{ runner.os }}.txt
 
-      - name: Get workflow job id
-        id: get-job-id
-        uses: ./.github/actions/get-workflow-job-id
-        if: always()
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-
       - name: Install PyTorch and run MPS tests
         id: test
         env:
@@ -110,14 +103,6 @@ jobs:
           NO_TEST_TIMEOUT: ${{ needs.filter.outputs.ci-no-test-timeout }}
           NO_TD: ${{ needs.filter.outputs.ci-no-td }}
           PIP_REQUIREMENTS_FILE: .github/requirements/pip-requirements-${{ runner.os }}.txt
-          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 }}
           REENABLED_ISSUES: ${{ needs.filter.outputs.reenabled-issues }}
         run: |
           # shellcheck disable=SC1090
@@ -159,6 +144,13 @@ jobs:
         run: |
           cat test/**/*_toprint.log || true
 
+      - name: Get workflow job id
+        id: get-job-id
+        uses: ./.github/actions/get-workflow-job-id
+        if: always()
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Upload test artifacts
         uses: ./.github/actions/upload-test-artifacts
         if: always() && steps.test.conclusion && steps.test.conclusion != 'skipped'

.github/workflows/_runner-determinator.yml

@@ -3,11 +3,6 @@ name: Check whether the workflow owner can use ARC runners
 on:
   workflow_call:
     inputs:
-      check_experiments:
-        required: false
-        type: string
-        description: |
-          List of experiments for this workfow. If not defined, all default experiments are included.
       triggering_actor:
         required: true
         type: string
@@ -48,7 +43,6 @@ jobs:
       ISSUE_NUMBER: ${{ inputs.issue_number }}
       TRIGGERING_ACTOR: ${{ inputs.triggering_actor }}
       ISSUE_OWNER: ${{ inputs.issue_owner }}
-      CHECK_EXPERIMENTS: ${{ inputs.check_experiments }}
     steps:
       # - name: Checkout PyTorch
       #   uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
@@ -65,10 +59,6 @@ jobs:
           cat <<EOF > runner_determinator.py
           # flake8: noqa: G004
 
-          # Note: Copies of this script in runner_determinator.py and _runner-determinator.yml
-          #       must be kept in sync. You can do it easily by running the following command:
-          #           python .github/scripts/update_runner_determinator.py
-
           """
           This runner determinator is used to determine which set of runners to run a
           GitHub job on. It uses the first comment of a GitHub issue (by default
@@ -104,8 +94,7 @@ jobs:
               experiments:
                 lf:
                   rollout_percent: 25
-                  all_branches: false
-                  default: true
+
               ---
 
               # Opt-ins:
@@ -123,7 +112,7 @@ jobs:
           import random
           from argparse import ArgumentParser
           from logging import LogRecord
-          from typing import Any, Dict, FrozenSet, Iterable, List, NamedTuple, Tuple
+          from typing import Any, Dict, Iterable, List, NamedTuple, Tuple
 
           import yaml
           from github import Auth, Github
@@ -149,12 +138,6 @@ jobs:
               rollout_perc: float = (
                   0  # Percentage of workflows to experiment on when user is not opted-in.
               )
-              all_branches: bool = (
-                  False  # If True, the experiment is also enabled on the exception branches
-              )
-              default: bool = (
-                  True  # If True, the experiment is enabled by default for all queries
-              )
 
               # Add more fields as needed
 
@@ -209,12 +192,6 @@ jobs:
                   f.write(f"{key}={value}\n")
 
 
-          def _str_comma_separated_to_set(value: str) -> FrozenSet[str]:
-              return frozenset(
-                  filter(lambda itm: itm != "", map(str.strip, value.strip(" \n\t").split(",")))
-              )
-
-
           def parse_args() -> Any:
               parser = ArgumentParser("Get dynamic rollout settings")
               parser.add_argument("--github-token", type=str, required=True, help="GitHub token")
@@ -249,13 +226,6 @@ jobs:
                   required=True,
                   help="Current GitHub ref type, branch or tag",
               )
-              parser.add_argument(
-                  "--eligible-experiments",
-                  type=_str_comma_separated_to_set,
-                  required=False,
-                  default="",
-                  help="comma separated list of experiments to check, if omitted all experiments marked with default=True are checked",
-              )
 
               return parser.parse_args()
 
@@ -301,7 +271,7 @@ jobs:
 
           def is_exception_branch(branch: str) -> bool:
               """
-              Branches that get opted out of experiments by default, until they're explicitly enabled.
+              Branches that get opted out of all experiments and should always use Meta runners
               """
               return branch.split("/")[0] in {"main", "nightly", "release", "landchecks"}
 
@@ -427,11 +397,7 @@ jobs:
 
 
           def get_runner_prefix(
-              rollout_state: str,
-              workflow_requestors: Iterable[str],
-              branch: str,
-              eligible_experiments: FrozenSet[str] = frozenset(),
-              is_canary: bool = False,
+              rollout_state: str, workflow_requestors: Iterable[str], is_canary: bool = False
           ) -> str:
               settings = parse_settings(rollout_state)
               user_optins = parse_users(rollout_state)
@@ -439,24 +405,7 @@ jobs:
               fleet_prefix = ""
               prefixes = []
               for experiment_name, experiment_settings in settings.experiments.items():
-                  if not experiment_settings.all_branches and is_exception_branch(branch):
-                      log.info(
-                          f"Branch {branch} is an exception branch. Not enabling experiment {experiment_name}."
-                      )
-                      continue
-
-                  if eligible_experiments:
-                      if experiment_name not in eligible_experiments:
-                          exp_list = ", ".join(eligible_experiments)
-                          log.info(
-                              f"Skipping experiment '{experiment_name}', as it is not in the eligible_experiments list: {exp_list}"
-                          )
-                          continue
-                  elif not experiment_settings.default:
-                      log.info(
-                          f"Skipping experiment '{experiment_name}', as it is not a default experiment"
-                      )
-                      continue
+                  enabled = False
 
                   # Is any workflow_requestor opted in to this experiment?
                   opted_in_users = [
@@ -465,13 +414,11 @@ jobs:
                       if is_user_opted_in(requestor, user_optins, experiment_name)
                   ]
 
-                  enabled = False
                   if opted_in_users:
                       log.info(
                           f"{', '.join(opted_in_users)} have opted into experiment {experiment_name}."
                       )
                       enabled = True
-
                   elif experiment_settings.rollout_perc:
                       # If no user is opted in, then we randomly enable the experiment based on the rollout percentage
                       if random.uniform(0, 100) <= experiment_settings.rollout_perc:
@@ -519,35 +466,35 @@ jobs:
           def main() -> None:
               args = parse_args()
 
-              runner_label_prefix = DEFAULT_LABEL_PREFIX
-
-              try:
-                  rollout_state = get_rollout_state_from_issue(
-                      args.github_token, args.github_issue_repo, args.github_issue
+              if args.github_ref_type == "branch" and is_exception_branch(args.github_branch):
+                  log.info(
+                      f"Exception branch: '{args.github_branch}', using Meta runners and no experiments."
                   )
+                  runner_label_prefix = DEFAULT_LABEL_PREFIX
+              else:
+                  try:
+                      rollout_state = get_rollout_state_from_issue(
+                          args.github_token, args.github_issue_repo, args.github_issue
+                      )
 
-                  username = get_potential_pr_author(
-                      args.github_token,
-                      args.github_repo,
-                      args.github_actor,
-                      args.github_ref_type,
-                      args.github_branch,
-                  )
+                      username = get_potential_pr_author(
+                          args.github_token,
+                          args.github_repo,
+                          args.github_actor,
+                          args.github_ref_type,
+                          args.github_branch,
+                      )
 
-                  is_canary = args.github_repo == "pytorch/pytorch-canary"
+                      is_canary = args.github_repo == "pytorch/pytorch-canary"
 
-                  runner_label_prefix = get_runner_prefix(
-                      rollout_state,
-                      (args.github_issue_owner, username),
-                      args.github_branch,
-                      args.eligible_experiments,
-                      is_canary,
-                  )
+                      runner_label_prefix = get_runner_prefix(
+                          rollout_state, (args.github_issue_owner, username), is_canary
+                      )
 
-              except Exception as e:
-                  log.error(
-                      f"Failed to get issue. Defaulting to Meta runners and no experiments. Exception: {e}"
-                  )
+                  except Exception as e:
+                      log.error(
+                          f"Failed to get issue. Defaulting to Meta runners and no experiments. Exception: {e}"
+                      )
 
               set_github_output(GH_OUTPUT_KEY_LABEL_TYPE, runner_label_prefix)
 
@@ -576,5 +523,4 @@ jobs:
             --github-actor "$TRIGGERING_ACTOR" \
             --github-issue-owner "$ISSUE_OWNER" \
             --github-ref-type "$curr_ref_type" \
-            --github-repo "$GITHUB_REPOSITORY" \
-            --eligible-experiments "$CHECK_EXPERIMENTS" \
+            --github-repo "$GITHUB_REPOSITORY"

.github/workflows/_win-test.yml

@@ -189,7 +189,7 @@ jobs:
         run: |
           pushd "${PYTORCH_FINAL_PACKAGE_DIR}"
           # shellcheck disable=SC2046,SC2102
-          python3 -mpip install $(echo *.whl)[opt-einsum,optree] optree==0.13.0
+          python3 -mpip install $(echo *.whl)[opt-einsum,optree] optree==0.12.1
           popd
 
           .ci/pytorch/win-test.sh

.github/workflows/build-conda-images.yml

@@ -32,7 +32,7 @@ concurrency:
 jobs:
   build-docker:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
-    runs-on: linux.9xlarge.ephemeral
+    runs-on: am2.linux.9xlarge.ephemeral
     strategy:
       matrix:
         cuda_version: ["11.8", "12.1", "12.4", "cpu"]
@@ -62,11 +62,5 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/conda/build.sh conda-builder${{ matrix.cuda_version == 'cpu' && ':' || ':cuda' }}${{matrix.cuda_version}}
+        run: |
+          .ci/docker/conda/build.sh conda-builder${{ matrix.cuda_version == 'cpu' && ':' || ':cuda' }}${{matrix.cuda_version}}

.github/workflows/build-libtorch-images.yml

@@ -31,7 +31,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -72,14 +72,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/libtorch/build.sh libtorch-cxx11-builder:cuda${{matrix.cuda_version}}
+        run: |
+          .ci/docker/libtorch/build.sh libtorch-cxx11-builder:cuda${{matrix.cuda_version}}
   build-docker-rocm:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -114,14 +108,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/libtorch/build.sh libtorch-cxx11-builder:rocm${{matrix.rocm_version}}
+        run: |
+          .ci/docker/libtorch/build.sh libtorch-cxx11-builder:rocm${{matrix.rocm_version}}
   build-docker-cpu:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -150,11 +138,5 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/libtorch/build.sh libtorch-cxx11-builder:cpu
+        run: |
+          .ci/docker/libtorch/build.sh libtorch-cxx11-builder:cpu

.github/workflows/build-manywheel-images.yml

@@ -35,7 +35,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -45,7 +45,7 @@ jobs:
   build-docker-cuda:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
-    runs-on: "${{ needs.get-label-type.outputs.label-type }}linux.9xlarge.ephemeral"
+    runs-on: "${{ needs.get-label-type.outputs.label-type }}am2.linux.9xlarge.ephemeral"
     strategy:
       matrix:
         cuda_version: ["12.4", "12.1", "11.8"]
@@ -78,14 +78,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux-builder:cuda${{matrix.cuda_version}}
+        run: |
+          .ci/docker/manywheel/build.sh manylinux-builder:cuda${{matrix.cuda_version}}
   # NOTE: manylinux_2_28 are still experimental, see https://github.com/pytorch/pytorch/issues/123649
   build-docker-cuda-manylinux_2_28:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
@@ -123,14 +117,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux2_28-builder:cuda${{matrix.cuda_version}}
+        run: |
+          .ci/docker/manywheel/build.sh manylinux2_28-builder:cuda${{matrix.cuda_version}}
   build-docker-cuda-aarch64:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -163,18 +151,12 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinuxaarch64-builder:cuda${{matrix.cuda_version}}
+        run: |
+          .ci/docker/manywheel/build.sh manylinuxaarch64-builder:cuda${{matrix.cuda_version}}
   build-docker-rocm:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
-    runs-on: "${{ needs.get-label-type.outputs.label-type }}linux.9xlarge.ephemeral"
+    runs-on: "${{ needs.get-label-type.outputs.label-type }}am2.linux.9xlarge.ephemeral"
     strategy:
       matrix:
         rocm_version: ["6.1", "6.2"]
@@ -205,18 +187,12 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux-builder:rocm${{matrix.rocm_version}}
+        run: |
+          .ci/docker/manywheel/build.sh manylinux-builder:rocm${{matrix.rocm_version}}
   build-docker-cpu:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
-    runs-on: "${{ needs.get-label-type.outputs.label-type }}linux.9xlarge.ephemeral"
+    runs-on: "${{ needs.get-label-type.outputs.label-type }}am2.linux.9xlarge.ephemeral"
     steps:
       - name: Checkout PyTorch
         uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
@@ -241,14 +217,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux-builder:cpu
+        run: |
+          .ci/docker/manywheel/build.sh manylinux-builder:cpu
   build-docker-cpu-manylinux_2_28:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -279,14 +249,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux2_28-builder:cpu
+        run: |
+          .ci/docker/manywheel/build.sh manylinux2_28-builder:cpu
   build-docker-cpu-aarch64:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -317,14 +281,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinuxaarch64-builder:cpu-aarch64
+        run: |
+          .ci/docker/manywheel/build.sh manylinuxaarch64-builder:cpu-aarch64
   build-docker-cpu-aarch64-2_28:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -358,14 +316,8 @@ jobs:
         env:
           DOCKER_TOKEN: ${{ secrets.DOCKER_TOKEN }}
           DOCKER_ID: ${{ secrets.DOCKER_ID }}
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux2_28_aarch64-builder:cpu-aarch64
+        run: |
+          .ci/docker/manywheel/build.sh manylinux2_28_aarch64-builder:cpu-aarch64
   build-docker-cpu-cxx11-abi:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -396,14 +348,8 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinuxcxx11-abi-builder:cpu-cxx11-abi
+        run: |
+          .ci/docker/manywheel/build.sh manylinuxcxx11-abi-builder:cpu-cxx11-abi
   build-docker-xpu:
     environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.event.ref, 'refs/tags/v')) && 'docker-build' || '' }}
     needs: get-label-type
@@ -434,11 +380,5 @@ jobs:
           fi
       - name: Build Docker Image
         if: env.WITH_PUSH == 'true'
-        uses: nick-fields/retry@v3.0.0
-        with:
-          shell: bash
-          timeout_minutes: 90
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            .ci/docker/manywheel/build.sh manylinux2_28-builder:xpu
+        run: |
+          .ci/docker/manywheel/build.sh manylinux2_28-builder:xpu

.github/workflows/build-triton-wheel.yml

@@ -29,7 +29,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -43,7 +43,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        py_vers: [ "3.9", "3.10", "3.11", "3.12" ]
+        py_vers: [ "3.8", "3.9", "3.10", "3.11", "3.12" ]
         device: ["cuda", "rocm", "xpu"]
         include:
           - device: "rocm"
@@ -91,6 +91,9 @@ jobs:
 
           # Determine python executable for given version
           case $PY_VERS in
+          3.8)
+            PYTHON_EXECUTABLE=/opt/python/cp38-cp38/bin/python
+            ;;
           3.9)
             PYTHON_EXECUTABLE=/opt/python/cp39-cp39/bin/python
             ;;
@@ -211,7 +214,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        py_vers: [ "3.9", "3.10", "3.11", "3.12" ]
+        py_vers: [ "3.8", "3.9", "3.10", "3.11", "3.12" ]
     timeout-minutes: 40
     env:
       DOCKER_IMAGE: pytorch/conda-builder:cpu

.github/workflows/check-labels.yml

@@ -30,9 +30,6 @@ concurrency:
 
 jobs:
   check-labels:
-    permissions:
-      contents: read
-      pull-requests: write
     name: Check labels
     if: github.repository_owner == 'pytorch'
     runs-on: linux.20_04.4x

.github/workflows/create_release.yml

@@ -18,7 +18,7 @@ on:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/docker-builds.yml

@@ -32,7 +32,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -67,7 +67,6 @@ jobs:
           pytorch-linux-jammy-py3.12-halide,
           pytorch-linux-jammy-xpu-2024.0-py3,
           pytorch-linux-jammy-py3-clang15-asan,
-          pytorch-linux-jammy-py3-clang18-asan,
           pytorch-linux-focal-py3-clang10-onnx,
           pytorch-linux-focal-linter,
           pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-linter,
@@ -79,9 +78,7 @@ jobs:
           - docker-image-name: pytorch-linux-jammy-aarch64-py3.10-gcc11-inductor-benchmarks
             runner: linux.arm64.m7g.4xlarge
             timeout-minutes: 600
-    # Docker uploads fail from LF runners, see https://github.com/pytorch/pytorch/pull/137358
-    # runs-on: "${{ needs.get-label-type.outputs.label-type }}${{ matrix.runner }}"
-    runs-on: "${{ matrix.runner }}"
+    runs-on: "${{ needs.get-label-type.outputs.label-type }}${{ matrix.runner }}"
     env:
       DOCKER_IMAGE_BASE: 308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/${{ matrix.docker-image-name }}
     steps:

.github/workflows/docker-release.yml

@@ -36,7 +36,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-linux-aarch64-binary-manywheel-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -60,7 +60,6 @@ jobs:
       DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
       use_split_build: False
       DESIRED_PYTHON: "3.9"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_9-cpu-aarch64
@@ -87,7 +86,6 @@ jobs:
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.2xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
     secrets:
@@ -132,7 +130,6 @@ jobs:
       DESIRED_DEVTOOLSET: cxx11-abi
       use_split_build: False
       DESIRED_PYTHON: "3.9"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_9-cuda-aarch64
@@ -180,7 +177,6 @@ jobs:
       DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
       use_split_build: False
       DESIRED_PYTHON: "3.10"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_10-cpu-aarch64
@@ -207,7 +203,6 @@ jobs:
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.2xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
     secrets:
@@ -252,7 +247,6 @@ jobs:
       DESIRED_DEVTOOLSET: cxx11-abi
       use_split_build: False
       DESIRED_PYTHON: "3.10"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_10-cuda-aarch64
@@ -300,7 +294,6 @@ jobs:
       DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
       use_split_build: False
       DESIRED_PYTHON: "3.11"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_11-cpu-aarch64
@@ -327,7 +320,6 @@ jobs:
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.2xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
     secrets:
@@ -372,7 +364,6 @@ jobs:
       DESIRED_DEVTOOLSET: cxx11-abi
       use_split_build: False
       DESIRED_PYTHON: "3.11"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_11-cuda-aarch64
@@ -420,7 +411,6 @@ jobs:
       DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
       use_split_build: False
       DESIRED_PYTHON: "3.12"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_12-cpu-aarch64
@@ -447,7 +437,6 @@ jobs:
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.2xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
     secrets:
@@ -492,7 +481,6 @@ jobs:
       DESIRED_DEVTOOLSET: cxx11-abi
       use_split_build: False
       DESIRED_PYTHON: "3.12"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       runs_on: linux.arm64.m7g.4xlarge.ephemeral
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_12-cuda-aarch64

.github/workflows/generated-linux-binary-conda-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-linux-binary-libtorch-cxx11-abi-main.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-linux-binary-libtorch-cxx11-abi-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-linux-binary-libtorch-pre-cxx11-main.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-linux-binary-libtorch-pre-cxx11-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-linux-binary-manywheel-main.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -153,7 +153,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_4-test:  # Testing

.github/workflows/generated-linux-binary-manywheel-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -343,7 +343,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_4-test:  # Testing
@@ -1029,7 +1029,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_10-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_10-cuda12_4-test:  # Testing
@@ -1785,7 +1785,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_11-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_11-cuda12_4-test:  # Testing
@@ -2471,7 +2471,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_12-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_12-cuda12_4-test:  # Testing
@@ -3157,7 +3157,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_13-cuda12_4
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_13-cuda12_4-test:  # Testing
@@ -3324,353 +3324,3 @@ jobs:
       conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
       conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
     uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cpu-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cpu
-      build_environment: linux-binary-manywheel
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cpu-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cpu-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cpu
-      build_environment: linux-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cpu-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cpu-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cpu-cxx11-abi-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu-cxx11-abi
-      GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
-      DESIRED_DEVTOOLSET: cxx11-abi
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cpu-cxx11-abi
-      build_environment: linux-binary-manywheel
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cpu-cxx11-abi-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cpu-cxx11-abi-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu-cxx11-abi
-      GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
-      DESIRED_DEVTOOLSET: cxx11-abi
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cpu-cxx11-abi
-      build_environment: linux-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cpu-cxx11-abi-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cpu-cxx11-abi-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu-cxx11-abi
-      GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
-      DESIRED_DEVTOOLSET: cxx11-abi
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cpu-cxx11-abi
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cuda11_8-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cuda11_8
-      build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | 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-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-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda11_8-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cuda11_8-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda11_8
-      build_environment: linux-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda11_8-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cuda11_8-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda11_8
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cuda12_1-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cuda12_1
-      build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | 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-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-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_1-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cuda12_1-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_1
-      build_environment: linux-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_1-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cuda12_1-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_1
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cuda12_4-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cuda12_4
-      build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_4-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cuda12_4-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_4
-      build_environment: linux-binary-manywheel
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_4-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cuda12_4-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: False
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_4
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml

.github/workflows/generated-linux-binary-manywheel-split-main.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -153,7 +153,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_4
       build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_4-test:  # Testing

.github/workflows/generated-linux-binary-manywheel-split-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -206,7 +206,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_4
       build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_4-test:  # Testing
@@ -486,7 +486,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_10-cuda12_4
       build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_10-cuda12_4-test:  # Testing
@@ -836,7 +836,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_11-cuda12_4
       build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_11-cuda12_4-test:  # Testing
@@ -1116,7 +1116,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_12-cuda12_4
       build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_12-cuda12_4-test:  # Testing
@@ -1396,7 +1396,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_13-cuda12_4
       build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_13-cuda12_4-test:  # Testing
@@ -1514,283 +1514,3 @@ jobs:
       conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
       conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
     uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cuda11_8-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cuda11_8
-      build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu11==11.8.89; platform_system == 'Linux' and platform_machine == 'x86_64' | 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-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-nvtx-cu11==11.8.86; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda11_8-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cuda11_8-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda11_8
-      build_environment: linux-binary-manywheel-split
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda11_8-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cuda11_8-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu118
-      GPU_ARCH_VERSION: 11.8
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda11_8
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cuda12_1-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cuda12_1
-      build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | 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-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-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_1-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cuda12_1-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_1
-      build_environment: linux-binary-manywheel-split
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_1-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cuda12_1-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu121
-      GPU_ARCH_VERSION: 12.1
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_1
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cuda12_4-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cuda12_4
-      build_environment: linux-binary-manywheel-split
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; 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==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; 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==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_4-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cuda12_4-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_4
-      build_environment: linux-binary-manywheel-split
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge.nvidia.gpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cuda12_4-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cuda12_4-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cu124
-      GPU_ARCH_VERSION: 12.4
-      GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cuda12_4
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml
-
-  manywheel-py3_13t-cpu-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    uses: ./.github/workflows/_binary-build-linux.yml
-    needs: get-label-type
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build_name: manywheel-py3_13t-cpu
-      build_environment: linux-binary-manywheel-split
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cpu-test:  # Testing
-    if: ${{ github.repository_owner == 'pytorch' }}
-    needs:
-      - manywheel-py3_13t-cpu-build
-      - get-label-type
-    uses: ./.github/workflows/_binary-test-linux.yml
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cpu
-      build_environment: linux-binary-manywheel-split
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runs_on: linux.4xlarge
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-  manywheel-py3_13t-cpu-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: manywheel-py3_13t-cpu-test
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: manywheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      use_split_build: True
-      DESIRED_PYTHON: "3.13t"
-      build_name: manywheel-py3_13t-cpu
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml

.github/workflows/generated-linux-s390x-binary-manywheel-nightly.yml

@@ -39,7 +39,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-macos-arm64-binary-wheel-nightly.yml

@@ -496,119 +496,3 @@ jobs:
       conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
       conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
     uses: ./.github/workflows/_binary-upload.yml
-  wheel-py3_13-cpu-build:
-    if: ${{ github.repository_owner == 'pytorch' }}
-    runs-on: macos-14-xlarge
-    timeout-minutes: 240
-    env:
-      PYTORCH_ROOT: ${{ github.workspace }}/pytorch
-      BUILDER_ROOT: ${{ github.workspace }}/builder
-      PACKAGE_TYPE: wheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      SKIP_ALL_TESTS: 1
-      DESIRED_PYTHON: "3.13"
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
-    steps:
-      # NOTE: These environment variables are put here so that they can be applied on every job equally
-      #       They are also here because setting them at a workflow level doesn't give us access to the
-      #       runner.temp variable, which we need.
-      - name: Populate binary env
-        shell: bash
-        run: |
-          # shellcheck disable=SC2129
-          echo "BINARY_ENV_FILE=${RUNNER_TEMP}/env" >> "${GITHUB_ENV}"
-          # shellcheck disable=SC2129
-          echo "PYTORCH_FINAL_PACKAGE_DIR=${RUNNER_TEMP}/artifacts" >> "${GITHUB_ENV}"
-          # shellcheck disable=SC2129
-          echo "MAC_PACKAGE_WORK_DIR=${RUNNER_TEMP}" >> "${GITHUB_ENV}"
-      - name: Install conda and dependencies
-        run: |
-          # Install conda, setup-miniconda messes with the path that messes with the ruby stuff we do later on
-          curl --retry 3 --retry-all-errors -o "${RUNNER_TEMP}/conda.sh" "https://repo.anaconda.com/miniconda/Miniconda3-py310_23.5.2-0-MacOSX-$(uname -m).sh"
-          chmod +x "${RUNNER_TEMP}/conda.sh"
-          /bin/bash "${RUNNER_TEMP}/conda.sh" -b -p "${RUNNER_TEMP}/anaconda"
-          echo "${RUNNER_TEMP}/anaconda/bin" >> "${GITHUB_PATH}"
-          if [ -d "/Applications/Xcode_14.3.1.app" ]; then
-            echo "DEVELOPER_DIR=/Applications/Xcode_14.3.1.app/Contents/Developer" >> "${GITHUB_ENV}"
-          elif [ -d "/Applications/Xcode_13.3.1.app" ]; then
-            echo "DEVELOPER_DIR=/Applications/Xcode_13.3.1.app/Contents/Developer" >> "${GITHUB_ENV}"
-          fi
-      - name: Checkout PyTorch
-        uses: malfet/checkout@silent-checkout
-        with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-          submodules: recursive
-          path: pytorch
-          quiet-checkout: true
-      - name: Clean PyTorch checkout
-        run: |
-          # Remove any artifacts from the previous checkouts
-          git clean -fxd
-        working-directory: pytorch
-      - name: Checkout pytorch/builder
-        uses: malfet/checkout@silent-checkout
-        with:
-          ref: main
-          submodules: recursive
-          repository: pytorch/builder
-          path: builder
-          quiet-checkout: true
-      - name: Clean pytorch/builder checkout
-        run: |
-          # Remove any artifacts from the previous checkouts
-          git clean -fxd
-        working-directory: builder
-      - name: Install sccache (only for non-forked PRs, and pushes to trunk)
-        uses: nick-fields/retry@v3.0.0
-        if: ${{ github.event_name == 'push' || github.event.pull_request.head.repo.full_name == github.repository }}
-        with:
-          timeout_minutes: 5
-          max_attempts: 3
-          retry_wait_seconds: 90
-          command: |
-            sudo curl --retry 3 --retry-all-errors https://s3.amazonaws.com/ossci-macos/sccache_v2.15 --output /usr/local/bin/sccache
-            sudo chmod +x /usr/local/bin/sccache
-            echo "SCCACHE_BUCKET=ossci-compiler-cache-circleci-v2" >> "${GITHUB_ENV}"
-      - name: Populate binary env
-        run: |
-          # shellcheck disable=SC1091
-          source "${RUNNER_TEMP}/anaconda/bin/activate"
-          "${PYTORCH_ROOT}/.circleci/scripts/binary_populate_env.sh"
-      - name: Build PyTorch binary
-        run: |
-          # shellcheck disable=SC1091
-          source "${RUNNER_TEMP}/anaconda/bin/activate"
-          "${PYTORCH_ROOT}/.circleci/scripts/binary_macos_build.sh"
-      - uses: actions/upload-artifact@v4.4.0
-        if: always()
-        with:
-          name: wheel-py3_13-cpu
-          retention-days: 14
-          if-no-files-found: error
-          path: "${{ env.PYTORCH_FINAL_PACKAGE_DIR }}"
-  wheel-py3_13-cpu-upload:  # Uploading
-    if: ${{ github.repository_owner == 'pytorch' }}
-    permissions:
-      id-token: write
-      contents: read
-    needs: wheel-py3_13-cpu-build
-    with:
-      PYTORCH_ROOT: /pytorch
-      BUILDER_ROOT: /builder
-      PACKAGE_TYPE: wheel
-      # TODO: This is a legacy variable that we eventually want to get rid of in
-      #       favor of GPU_ARCH_VERSION
-      DESIRED_CUDA: cpu
-      GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
-      DESIRED_PYTHON: "3.13"
-      build_name: wheel-py3_13-cpu
-      use_s3: False
-    secrets:
-      github-token: ${{ secrets.GITHUB_TOKEN }}
-      conda-pytorchbot-token: ${{ secrets.CONDA_PYTORCHBOT_TOKEN }}
-      conda-pytorchbot-token-test: ${{ secrets.CONDA_PYTORCHBOT_TOKEN_TEST }}
-    uses: ./.github/workflows/_binary-upload.yml

.github/workflows/generated-windows-binary-conda-nightly.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-windows-binary-libtorch-debug-main.yml

@@ -27,7 +27,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-windows-binary-libtorch-debug-nightly.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-windows-binary-libtorch-release-main.yml

@@ -27,7 +27,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-windows-binary-libtorch-release-nightly.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/generated-windows-binary-wheel-nightly.yml

@@ -34,7 +34,7 @@ concurrency:
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/inductor-cu124.yml

@@ -20,7 +20,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -56,7 +56,7 @@ jobs:
           { config: "aot_inductor_timm", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
           { config: "aot_inductor_torchbench", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
           { config: "aot_inductor_torchbench", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
-          { config: "inductor_cpp_wrapper", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
+          { config: "inductor_cpp_wrapper_abi_compatible", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
         ]}
     secrets:
       HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}

.github/workflows/inductor-micro-benchmark.yml

@@ -18,7 +18,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/inductor-perf-compare.yml

@@ -15,7 +15,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/inductor-perf-test-nightly-a10g.yml

@@ -70,7 +70,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}

.github/workflows/inductor-perf-test-nightly-aarch64.yml

@@ -5,7 +5,9 @@ on:
     # - cron: 0 7 * * 1-6
     # - cron: 0 7 * * 0
     # Does not perform max_autotune on CPU, so skip the weekly run setup
-    - cron: 0 7 * * *
+    # Run 6 times everyday to see if perf instablity can be reproduced
+    # Will change this back
+    - cron: 0 */4 * * *
   # NB: GitHub has an upper limit of 10 inputs here
   workflow_dispatch:
     inputs:
@@ -50,7 +52,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
@@ -114,7 +116,7 @@ jobs:
     name: linux-jammy-aarch64-py3.10-inductor
     uses: ./.github/workflows/_linux-test.yml
     needs: linux-jammy-aarch64-py3_10-inductor-build
-    if: github.event.schedule == '0 7 * * *'
+    if: github.event.schedule == '0 */4 * * *'
     with:
       build-environment: linux-jammy-aarch64-py3.10
       # Turn off dynamic-shapes and aotinductor tests for now, to have faster iteration for debugging perf instability.

.github/workflows/inductor-perf-test-nightly-x86.yml

@@ -50,7 +50,7 @@ permissions: read-all
 jobs:
   get-label-type:
     name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
+    uses: ./.github/workflows/_runner-determinator.yml
     with:
       triggering_actor: ${{ github.triggering_actor }}
       issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}