[ONNX] Support float4 (#151069)

- Support exporting float4 models (note: currently we use IR version 10 universally in the exporter, which does not include float 4 support. Eventually when onnx runtime and the ecosystem moves to support the new IR version 11 we should bump our version to 11 in the exporter as well)
- The shape of the type is set according to https://github.com/pytorch/pytorch/pull/148791#discussion_r2038704986 (added last dim with size 2)
- Use ml_dtypes types when converting to numpy for consistency with ONNX IR

Fix https://github.com/pytorch/pytorch/issues/150202

Pull Request resolved: https://github.com/pytorch/pytorch/pull/151069
Approved by: https://github.com/titaiwangms

.ci/aarch64_linux/aarch64_ci_build.sh

@@ -27,7 +27,6 @@ if [ "$DESIRED_CUDA" = "cpu" ]; then
     USE_PRIORITIZED_TEXT_FOR_LD=1 python /pytorch/.ci/aarch64_linux/aarch64_wheel_ci_build.py --enable-mkldnn
 else
     echo "BASE_CUDA_VERSION is set to: $DESIRED_CUDA"
-    export USE_SYSTEM_NCCL=1
     #USE_PRIORITIZED_TEXT_FOR_LD for enable linker script optimization https://github.com/pytorch/pytorch/pull/121975/files
     USE_PRIORITIZED_TEXT_FOR_LD=1 python /pytorch/.ci/aarch64_linux/aarch64_wheel_ci_build.py --enable-mkldnn --enable-cuda
 fi

.ci/aarch64_linux/aarch64_wheel_ci_build.py

@@ -31,47 +31,33 @@ def build_ArmComputeLibrary() -> None:
         "build=native",
     ]
     acl_install_dir = "/acl"
-    acl_checkout_dir = os.getenv("ACL_SOURCE_DIR", "ComputeLibrary")
-    if os.path.isdir(acl_install_dir):
-        shutil.rmtree(acl_install_dir)
-    if not os.path.isdir(acl_checkout_dir) or not len(os.listdir(acl_checkout_dir)):
-        check_call(
-            [
-                "git",
-                "clone",
-                "https://github.com/ARM-software/ComputeLibrary.git",
-                "-b",
-                "v25.02",
-                "--depth",
-                "1",
-                "--shallow-submodules",
-            ]
-        )
+    acl_checkout_dir = "ComputeLibrary"
+    os.makedirs(acl_install_dir)
+    check_call(
+        [
+            "git",
+            "clone",
+            "https://github.com/ARM-software/ComputeLibrary.git",
+            "-b",
+            "v25.02",
+            "--depth",
+            "1",
+            "--shallow-submodules",
+        ]
+    )
 
     check_call(
-        ["scons", "Werror=1", f"-j{os.cpu_count()}"] + acl_build_flags,
+        ["scons", "Werror=1", "-j8", f"build_dir=/{acl_install_dir}/build"]
+        + acl_build_flags,
         cwd=acl_checkout_dir,
     )
-    for d in ["arm_compute", "include", "utils", "support", "src", "build"]:
+    for d in ["arm_compute", "include", "utils", "support", "src"]:
         shutil.copytree(f"{acl_checkout_dir}/{d}", f"{acl_install_dir}/{d}")
 
 
-def replace_tag(filename) -> None:
-    with open(filename) as f:
-        lines = f.readlines()
-    for i, line in enumerate(lines):
-        if line.startswith("Tag:"):
-            lines[i] = line.replace("-linux_", "-manylinux_2_28_")
-            print(f"Updated tag from {line} to {lines[i]}")
-            break
-
-    with open(filename, "w") as f:
-        f.writelines(lines)
-
-
-def package_cuda_wheel(wheel_path, desired_cuda) -> None:
+def update_wheel(wheel_path, desired_cuda) -> None:
     """
-    Package the cuda wheel libraries
+    Update the cuda wheel libraries
     """
     folder = os.path.dirname(wheel_path)
     wheelname = os.path.basename(wheel_path)
@@ -102,19 +88,30 @@ def package_cuda_wheel(wheel_path, desired_cuda) -> None:
         "/usr/lib64/libgfortran.so.5",
         "/acl/build/libarm_compute.so",
         "/acl/build/libarm_compute_graph.so",
-        "/usr/local/lib/libnvpl_lapack_lp64_gomp.so.0",
-        "/usr/local/lib/libnvpl_blas_lp64_gomp.so.0",
-        "/usr/local/lib/libnvpl_lapack_core.so.0",
-        "/usr/local/lib/libnvpl_blas_core.so.0",
     ]
-
-    if "128" in desired_cuda:
+    if enable_cuda:
         libs_to_copy += [
-            "/usr/local/cuda/lib64/libnvrtc-builtins.so.12.8",
-            "/usr/local/cuda/lib64/libcufile.so.0",
-            "/usr/local/cuda/lib64/libcufile_rdma.so.1",
+            "/usr/local/lib/libnvpl_lapack_lp64_gomp.so.0",
+            "/usr/local/lib/libnvpl_blas_lp64_gomp.so.0",
+            "/usr/local/lib/libnvpl_lapack_core.so.0",
+            "/usr/local/lib/libnvpl_blas_core.so.0",
+        ]
+        if "126" in desired_cuda:
+            libs_to_copy += [
+                "/usr/local/cuda/lib64/libnvrtc-builtins.so.12.6",
+                "/usr/local/cuda/lib64/libcufile.so.0",
+                "/usr/local/cuda/lib64/libcufile_rdma.so.1",
+            ]
+        elif "128" in desired_cuda:
+            libs_to_copy += [
+                "/usr/local/cuda/lib64/libnvrtc-builtins.so.12.8",
+                "/usr/local/cuda/lib64/libcufile.so.0",
+                "/usr/local/cuda/lib64/libcufile_rdma.so.1",
+            ]
+    else:
+        libs_to_copy += [
+            "/opt/OpenBLAS/lib/libopenblas.so.0",
         ]
-
     # Copy libraries to unzipped_folder/a/lib
     for lib_path in libs_to_copy:
         lib_name = os.path.basename(lib_path)
@@ -123,13 +120,6 @@ def package_cuda_wheel(wheel_path, desired_cuda) -> None:
             f"cd {folder}/tmp/torch/lib/; "
             f"patchelf --set-rpath '$ORIGIN' --force-rpath {folder}/tmp/torch/lib/{lib_name}"
         )
-
-    # Make sure the wheel is tagged with manylinux_2_28
-    for f in os.scandir(f"{folder}/tmp/"):
-        if f.is_dir() and f.name.endswith(".dist-info"):
-            replace_tag(f"{f.path}/WHEEL")
-            break
-
     os.mkdir(f"{folder}/cuda_wheel")
     os.system(f"cd {folder}/tmp/; zip -r {folder}/cuda_wheel/{wheelname} *")
     shutil.move(
@@ -204,10 +194,8 @@ if __name__ == "__main__":
     ).decode()
 
     print("Building PyTorch wheel")
-    build_vars = "CMAKE_SHARED_LINKER_FLAGS=-Wl,-z,max-page-size=0x10000 "
-    # MAX_JOB=5 is not required for CPU backend (see commit 465d98b)
-    if enable_cuda:
-        build_vars = "MAX_JOBS=5 " + build_vars
+    build_vars = "MAX_JOBS=5 CMAKE_SHARED_LINKER_FLAGS=-Wl,-z,max-page-size=0x10000 "
+    os.system("cd /pytorch; python setup.py clean")
 
     override_package_version = os.getenv("OVERRIDE_PACKAGE_VERSION")
     desired_cuda = os.getenv("DESIRED_CUDA")
@@ -254,6 +242,6 @@ if __name__ == "__main__":
         print("Updating Cuda Dependency")
         filename = os.listdir("/pytorch/dist/")
         wheel_path = f"/pytorch/dist/{filename[0]}"
-        package_cuda_wheel(wheel_path, desired_cuda)
+        update_wheel(wheel_path, desired_cuda)
     pytorch_wheel_name = complete_wheel("/pytorch/")
     print(f"Build Complete. Created {pytorch_wheel_name}..")

.ci/docker/almalinux/Dockerfile

@@ -1,7 +1,7 @@
 ARG CUDA_VERSION=12.4
 ARG BASE_TARGET=cuda${CUDA_VERSION}
 ARG ROCM_IMAGE=rocm/dev-almalinux-8:6.3-complete
-FROM amd64/almalinux:8.10-20250519 as base
+FROM amd64/almalinux:8 as base
 
 ENV LC_ALL en_US.UTF-8
 ENV LANG en_US.UTF-8
@@ -11,8 +11,6 @@ ARG DEVTOOLSET_VERSION=11
 
 RUN yum -y update
 RUN yum -y install epel-release
-# install glibc-langpack-en make sure en_US.UTF-8 locale is available
-RUN yum -y install glibc-langpack-en
 RUN yum install -y sudo wget curl perl util-linux xz bzip2 git patch which perl zlib-devel openssl-devel yum-utils autoconf automake make gcc-toolset-${DEVTOOLSET_VERSION}-toolchain
 # Just add everything as a safe.directory for git since these will be used in multiple places with git
 RUN git config --global --add safe.directory '*'
@@ -78,7 +76,7 @@ RUN bash ./install_mnist.sh
 FROM base as all_cuda
 COPY --from=cuda11.8  /usr/local/cuda-11.8 /usr/local/cuda-11.8
 COPY --from=cuda12.6  /usr/local/cuda-12.6 /usr/local/cuda-12.6
-COPY --from=cuda12.8  /usr/local/cuda-12.8 /usr/local/cuda-12.8
+COPY --from=cuda12.4  /usr/local/cuda-12.8 /usr/local/cuda-12.8
 
 # Final step
 FROM ${BASE_TARGET} as final

.ci/docker/build.sh

@@ -50,21 +50,30 @@ if [[ "$image" == *xla* ]]; then
   exit 0
 fi
 
-if [[ "$image" == *-jammy* ]]; then
+if [[ "$image" == *-focal* ]]; then
+  UBUNTU_VERSION=20.04
+elif [[ "$image" == *-jammy* ]]; then
   UBUNTU_VERSION=22.04
 elif [[ "$image" == *ubuntu* ]]; then
   extract_version_from_image_name ubuntu UBUNTU_VERSION
+elif [[ "$image" == *centos* ]]; then
+  extract_version_from_image_name centos CENTOS_VERSION
 fi
 
 if [ -n "${UBUNTU_VERSION}" ]; then
   OS="ubuntu"
+elif [ -n "${CENTOS_VERSION}" ]; then
+  OS="centos"
 else
   echo "Unable to derive operating system base..."
   exit 1
 fi
 
 DOCKERFILE="${OS}/Dockerfile"
-if [[ "$image" == *rocm* ]]; then
+# When using ubuntu - 22.04, start from Ubuntu docker image, instead of nvidia/cuda docker image.
+if [[ "$image" == *cuda* && "$UBUNTU_VERSION" != "22.04" ]]; then
+  DOCKERFILE="${OS}-cuda/Dockerfile"
+elif [[ "$image" == *rocm* ]]; then
   DOCKERFILE="${OS}-rocm/Dockerfile"
 elif [[ "$image" == *xpu* ]]; then
   DOCKERFILE="${OS}-xpu/Dockerfile"
@@ -89,7 +98,7 @@ tag=$(echo $image | awk -F':' '{print $2}')
 # configuration, so we hardcode everything here rather than do it
 # from scratch
 case "$tag" in
-  pytorch-linux-jammy-cuda12.6-cudnn9-py3-gcc11)
+  pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc11)
     CUDA_VERSION=12.6.3
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.10
@@ -100,8 +109,8 @@ case "$tag" in
     UCC_COMMIT=${_UCC_COMMIT}
     TRITON=yes
     ;;
-  pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.8.1
+  pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9-inductor-benchmarks)
+    CUDA_VERSION=12.4.1
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
@@ -112,8 +121,8 @@ case "$tag" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.8-cudnn9-py3.12-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.8.1
+  pytorch-linux-focal-cuda12.4-cudnn9-py3.12-gcc9-inductor-benchmarks)
+    CUDA_VERSION=12.4.1
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.12
     GCC_VERSION=9
@@ -124,8 +133,8 @@ case "$tag" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.8-cudnn9-py3.13-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.8.1
+  pytorch-linux-focal-cuda12.4-cudnn9-py3.13-gcc9-inductor-benchmarks)
+    CUDA_VERSION=12.4.1
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.13
     GCC_VERSION=9
@@ -136,7 +145,7 @@ case "$tag" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.6-cudnn9-py3-gcc9)
+  pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9)
     CUDA_VERSION=12.6.3
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.10
@@ -147,8 +156,8 @@ case "$tag" in
     UCC_COMMIT=${_UCC_COMMIT}
     TRITON=yes
     ;;
-  pytorch-linux-jammy-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.6
+  pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks)
+    CUDA_VERSION=12.6.3
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
@@ -159,8 +168,8 @@ case "$tag" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.6-cudnn9-py3.12-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.6
+  pytorch-linux-focal-cuda12.6-cudnn9-py3.12-gcc9-inductor-benchmarks)
+    CUDA_VERSION=12.6.3
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.12
     GCC_VERSION=9
@@ -171,8 +180,8 @@ case "$tag" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.6-cudnn9-py3.13-gcc9-inductor-benchmarks)
-    CUDA_VERSION=12.6
+  pytorch-linux-focal-cuda12.6-cudnn9-py3.13-gcc9-inductor-benchmarks)
+    CUDA_VERSION=12.6.3
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.13
     GCC_VERSION=9
@@ -183,8 +192,8 @@ case "$tag" in
     TRITON=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9)
-    CUDA_VERSION=12.8.1
+  pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9)
+    CUDA_VERSION=11.8.0
     CUDNN_VERSION=9
     ANACONDA_PYTHON_VERSION=3.10
     GCC_VERSION=9
@@ -194,25 +203,25 @@ case "$tag" in
     UCC_COMMIT=${_UCC_COMMIT}
     TRITON=yes
     ;;
-  pytorch-linux-jammy-py3-clang12-onnx)
+  pytorch-linux-focal-py3-clang10-onnx)
     ANACONDA_PYTHON_VERSION=3.9
-    CLANG_VERSION=12
+    CLANG_VERSION=10
     VISION=yes
     ONNX=yes
     ;;
-  pytorch-linux-jammy-py3.9-clang12)
+  pytorch-linux-focal-py3.9-clang10)
     ANACONDA_PYTHON_VERSION=3.9
-    CLANG_VERSION=12
+    CLANG_VERSION=10
     VISION=yes
     TRITON=yes
     ;;
-  pytorch-linux-jammy-py3.11-clang12)
+  pytorch-linux-focal-py3.11-clang10)
     ANACONDA_PYTHON_VERSION=3.11
-    CLANG_VERSION=12
+    CLANG_VERSION=10
     VISION=yes
     TRITON=yes
     ;;
-  pytorch-linux-jammy-py3.9-gcc9)
+  pytorch-linux-focal-py3.9-gcc9)
     ANACONDA_PYTHON_VERSION=3.9
     GCC_VERSION=9
     VISION=yes
@@ -267,9 +276,9 @@ case "$tag" in
     DOCS=yes
     INDUCTOR_BENCHMARKS=yes
     ;;
-  pytorch-linux-jammy-cuda12.8-cudnn9-py3.9-clang12)
+  pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-clang12)
     ANACONDA_PYTHON_VERSION=3.9
-    CUDA_VERSION=12.8.1
+    CUDA_VERSION=11.8
     CUDNN_VERSION=9
     CLANG_VERSION=12
     VISION=yes
@@ -318,15 +327,15 @@ case "$tag" in
     GCC_VERSION=11
     TRITON_CPU=yes
     ;;
-  pytorch-linux-jammy-linter)
+  pytorch-linux-focal-linter)
     # TODO: Use 3.9 here because of this issue https://github.com/python/mypy/issues/13627.
     # We will need to update mypy version eventually, but that's for another day. The task
     # would be to upgrade mypy to 1.0.0 with Python 3.11
     PYTHON_VERSION=3.9
     ;;
-  pytorch-linux-jammy-cuda12.8-cudnn9-py3.9-linter)
+  pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-linter)
     PYTHON_VERSION=3.9
-    CUDA_VERSION=12.8.1
+    CUDA_VERSION=11.8
     ;;
   pytorch-linux-jammy-aarch64-py3.10-gcc11)
     ANACONDA_PYTHON_VERSION=3.10
@@ -385,6 +394,14 @@ esac
 
 tmp_tag=$(basename "$(mktemp -u)" | tr '[:upper:]' '[:lower:]')
 
+#when using cudnn version 8 install it separately from cuda
+if [[ "$image" == *cuda*  && ${OS} == "ubuntu" ]]; then
+  IMAGE_NAME="nvidia/cuda:${CUDA_VERSION}-cudnn${CUDNN_VERSION}-devel-ubuntu${UBUNTU_VERSION}"
+  if [[ ${CUDNN_VERSION} == 9 ]]; then
+    IMAGE_NAME="nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}"
+  fi
+fi
+
 no_cache_flag=""
 progress_flag=""
 # Do not use cache and progress=plain when in CI
@@ -401,6 +418,7 @@ docker build \
        --build-arg "LLVMDEV=${LLVMDEV:-}" \
        --build-arg "VISION=${VISION:-}" \
        --build-arg "UBUNTU_VERSION=${UBUNTU_VERSION}" \
+       --build-arg "CENTOS_VERSION=${CENTOS_VERSION}" \
        --build-arg "DEVTOOLSET_VERSION=${DEVTOOLSET_VERSION}" \
        --build-arg "GLIBC_VERSION=${GLIBC_VERSION}" \
        --build-arg "CLANG_VERSION=${CLANG_VERSION}" \

.ci/docker/ci_commit_pins/executorch.txt

@@ -1 +1 @@
-f50bfa92602b45dca884a9e511e5d9ddbe8ba314
+b173722085b3f555d6ba4533d6bbaddfd7c71144

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

@@ -1 +1 @@
-b0e26b7359c147b8aa0af686c20510fb9b15990a
+0bcc8265e677e5321606a3311bf71470f14456a8

.ci/docker/ci_commit_pins/triton.txt

@@ -1 +1 @@
-c8757738a7418249896224430ce84888e8ecdd79
+96316ce50fade7e209553aba4898cd9b82aab83b

.ci/docker/common/install_base.sh

@@ -36,6 +36,8 @@ install_ubuntu() {
   # See https://github.com/pytorch/pytorch/issues/144768
   if [[ "$UBUNTU_VERSION" == "20.04"* && "$CUDA_VERSION" == "11.8"* ]]; then
     maybe_libnccl_dev="libnccl2=2.15.5-1+cuda11.8 libnccl-dev=2.15.5-1+cuda11.8 --allow-downgrades --allow-change-held-packages"
+  elif [[ "$UBUNTU_VERSION" == "20.04"* && "$CUDA_VERSION" == "12.4"* ]]; then
+    maybe_libnccl_dev="libnccl2=2.26.2-1+cuda12.4 libnccl-dev=2.26.2-1+cuda12.4 --allow-downgrades --allow-change-held-packages"
   else
     maybe_libnccl_dev=""
   fi

.ci/docker/common/install_cuda.sh

@@ -54,9 +54,23 @@ function install_118 {
     ldconfig
 }
 
+function install_124 {
+  CUDNN_VERSION=9.1.0.70
+  echo "Installing CUDA 12.4.1 and cuDNN ${CUDNN_VERSION} and NCCL and cuSparseLt-0.6.2"
+  install_cuda 12.4.1 cuda_12.4.1_550.54.15_linux
+
+  install_cudnn 12 $CUDNN_VERSION
+
+  CUDA_VERSION=12.4 bash install_nccl.sh
+
+  CUDA_VERSION=12.4 bash install_cusparselt.sh
+
+  ldconfig
+}
+
 function install_126 {
   CUDNN_VERSION=9.5.1.17
-  echo "Installing CUDA 12.6.3 and cuDNN ${CUDNN_VERSION} and NCCL and cuSparseLt-0.7.1"
+  echo "Installing CUDA 12.6.3 and cuDNN ${CUDNN_VERSION} and NCCL and cuSparseLt-0.6.3"
   install_cuda 12.6.3 cuda_12.6.3_560.35.05_linux
 
   install_cudnn 12 $CUDNN_VERSION
@@ -99,6 +113,40 @@ function prune_118 {
     rm -rf $CUDA_BASE/libnvvp $CUDA_BASE/nsightee_plugins $CUDA_BASE/nsight-compute-2022.3.0 $CUDA_BASE/nsight-systems-2022.4.2/
 }
 
+function prune_124 {
+  echo "Pruning CUDA 12.4"
+  #####################################################################################
+  # CUDA 12.4 prune static libs
+  #####################################################################################
+  export NVPRUNE="/usr/local/cuda-12.4/bin/nvprune"
+  export CUDA_LIB_DIR="/usr/local/cuda-12.4/lib64"
+
+  export GENCODE="-gencode arch=compute_50,code=sm_50 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_90,code=sm_90"
+  export GENCODE_CUDNN="-gencode arch=compute_50,code=sm_50 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_90,code=sm_90"
+
+  if [[ -n "$OVERRIDE_GENCODE" ]]; then
+      export GENCODE=$OVERRIDE_GENCODE
+  fi
+  if [[ -n "$OVERRIDE_GENCODE_CUDNN" ]]; then
+      export GENCODE_CUDNN=$OVERRIDE_GENCODE_CUDNN
+  fi
+
+  # all CUDA libs except CuDNN and CuBLAS
+  ls $CUDA_LIB_DIR/ | grep "\.a" | grep -v "culibos" | grep -v "cudart" | grep -v "cudnn" | grep -v "cublas" | grep -v "metis"  \
+      | xargs -I {} bash -c \
+                "echo {} && $NVPRUNE $GENCODE $CUDA_LIB_DIR/{} -o $CUDA_LIB_DIR/{}"
+
+  # prune CuDNN and CuBLAS
+  $NVPRUNE $GENCODE_CUDNN $CUDA_LIB_DIR/libcublas_static.a -o $CUDA_LIB_DIR/libcublas_static.a
+  $NVPRUNE $GENCODE_CUDNN $CUDA_LIB_DIR/libcublasLt_static.a -o $CUDA_LIB_DIR/libcublasLt_static.a
+
+  #####################################################################################
+  # CUDA 12.4 prune visual tools
+  #####################################################################################
+  export CUDA_BASE="/usr/local/cuda-12.4/"
+  rm -rf $CUDA_BASE/libnvvp $CUDA_BASE/nsightee_plugins $CUDA_BASE/nsight-compute-2024.1.0 $CUDA_BASE/nsight-systems-2023.4.4/
+}
+
 function prune_126 {
   echo "Pruning CUDA 12.6"
   #####################################################################################
@@ -135,9 +183,9 @@ function prune_126 {
 
 function install_128 {
   CUDNN_VERSION=9.8.0.87
-  echo "Installing CUDA 12.8.1 and cuDNN ${CUDNN_VERSION} and NCCL and cuSparseLt-0.7.1"
-  # install CUDA 12.8.1 in the same container
-  install_cuda 12.8.1 cuda_12.8.1_570.124.06_linux
+  echo "Installing CUDA 12.8.0 and cuDNN ${CUDNN_VERSION} and NCCL and cuSparseLt-0.6.3"
+  # install CUDA 12.8.0 in the same container
+  install_cuda 12.8.0 cuda_12.8.0_570.86.10_linux
 
   # cuDNN license: https://developer.nvidia.com/cudnn/license_agreement
   install_cudnn 12 $CUDNN_VERSION
@@ -155,9 +203,11 @@ do
     case "$1" in
     11.8) install_118; prune_118
         ;;
-    12.6|12.6.*) install_126; prune_126
+    12.4) install_124; prune_124
         ;;
-    12.8|12.8.*) install_128;
+    12.6) install_126; prune_126
+        ;;
+    12.8) install_128;
         ;;
     *) echo "bad argument $1"; exit 1
         ;;

.ci/docker/common/install_cudnn.sh

@@ -8,6 +8,8 @@ if [[ -n "${CUDNN_VERSION}" ]]; then
         CUDNN_NAME="cudnn-linux-x86_64-9.8.0.87_cuda12-archive"
     elif [[ ${CUDA_VERSION:0:4} == "12.6" ]]; then
         CUDNN_NAME="cudnn-linux-x86_64-9.5.1.17_cuda12-archive"
+    elif [[ ${CUDA_VERSION:0:2} == "12" ]]; then
+        CUDNN_NAME="cudnn-linux-x86_64-9.1.0.70_cuda12-archive"
     elif [[ ${CUDA_VERSION:0:2} == "11" ]]; then
         CUDNN_NAME="cudnn-linux-x86_64-9.1.0.70_cuda11-archive"
     else

.ci/docker/common/install_cusparselt.sh

@@ -11,7 +11,15 @@ if [[ ${CUDA_VERSION:0:4} =~ ^12\.[5-8]$ ]]; then
     if [ ${TARGETARCH} = 'amd64' ] || [ "${TARGETARCH}" = 'x86_64' ]; then
         arch_path='x86_64'
     fi
-    CUSPARSELT_NAME="libcusparse_lt-linux-${arch_path}-0.7.1.0-archive"
+    CUSPARSELT_NAME="libcusparse_lt-linux-${arch_path}-0.6.3.2-archive"
+    curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-${arch_path}/${CUSPARSELT_NAME}.tar.xz
+elif [[ ${CUDA_VERSION:0:4} == "12.4" ]]; then
+    arch_path='sbsa'
+    export TARGETARCH=${TARGETARCH:-$(uname -m)}
+    if [ ${TARGETARCH} = 'amd64' ] || [ "${TARGETARCH}" = 'x86_64' ]; then
+        arch_path='x86_64'
+    fi
+    CUSPARSELT_NAME="libcusparse_lt-linux-${arch_path}-0.6.2.3-archive"
     curl --retry 3 -OLs https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-${arch_path}/${CUSPARSELT_NAME}.tar.xz
 elif [[ ${CUDA_VERSION:0:4} == "11.8" ]]; then
     CUSPARSELT_NAME="libcusparse_lt-linux-x86_64-0.4.0.7-archive"

.ci/docker/common/install_onnx.sh

@@ -8,6 +8,16 @@ retry () {
     "$@" || (sleep 10 && "$@") || (sleep 20 && "$@") || (sleep 40 && "$@")
 }
 
+# A bunch of custom pip dependencies for ONNX
+pip_install \
+  beartype==0.15.0 \
+  filelock==3.9.0 \
+  flatbuffers==2.0 \
+  mock==5.0.1 \
+  ninja==1.10.2 \
+  networkx==2.5 \
+  numpy==1.24.2
+
 # ONNXRuntime should be installed before installing
 # onnx-weekly. Otherwise, onnx-weekly could be
 # overwritten by onnx.
@@ -19,8 +29,12 @@ pip_install \
   transformers==4.36.2
 
 pip_install coloredlogs packaging
+
 pip_install onnxruntime==1.18.1
-pip_install onnxscript==0.3.0
+pip_install onnx==1.17.0
+pip_install onnxscript==0.2.2 --no-deps
+# required by onnxscript
+pip_install ml_dtypes
 
 # Cache the transformers model to be used later by ONNX tests. We need to run the transformers
 # package to download the model. By default, the model is cached at ~/.cache/huggingface/hub/

.ci/docker/common/install_openblas.sh

@@ -4,7 +4,8 @@
 set -ex
 
 cd /
-git clone https://github.com/OpenMathLib/OpenBLAS.git -b "${OPENBLAS_VERSION:-v0.3.29}" --depth 1 --shallow-submodules
+git clone https://github.com/OpenMathLib/OpenBLAS.git -b v0.3.29 --depth 1 --shallow-submodules
+
 
 OPENBLAS_BUILD_FLAGS="
 NUM_THREADS=128

.ci/docker/common/install_triton.sh

@@ -51,12 +51,7 @@ as_jenkins git clone --recursive ${TRITON_REPO} triton
 cd triton
 as_jenkins git checkout ${TRITON_PINNED_COMMIT}
 as_jenkins git submodule update --init --recursive
-
-# Old versions of python have setup.py in ./python; newer versions have it in ./
-if [ ! -f setup.py ]; then
-  cd python
-fi
-
+cd python
 pip_install pybind11==2.13.6
 
 # TODO: remove patch setup.py once we have a proper fix for https://github.com/triton-lang/triton/issues/4527

.ci/docker/manywheel/Dockerfile_2_28_aarch64

@@ -58,7 +58,6 @@ RUN git config --global --add safe.directory "*"
 
 FROM base as openblas
 # Install openblas
-ARG OPENBLAS_VERSION
 ADD ./common/install_openblas.sh install_openblas.sh
 RUN bash ./install_openblas.sh && rm install_openblas.sh
 

.ci/docker/manywheel/Dockerfile_s390x

@@ -5,9 +5,7 @@ ENV LC_ALL=C.UTF-8
 ENV LANG=C.UTF-8
 ENV LANGUAGE=C.UTF-8
 
-# there is a bugfix in gcc >= 14 for precompiled headers and s390x vectorization interaction.
-# with earlier gcc versions test/inductor/test_cpu_cpp_wrapper.py will fail.
-ARG DEVTOOLSET_VERSION=14
+ARG DEVTOOLSET_VERSION=13
 # Installed needed OS packages. This is to support all
 # the binary builds (torch, vision, audio, text, data)
 RUN yum -y install epel-release
@@ -60,8 +58,7 @@ RUN yum install -y \
   libxslt-devel \
   libxml2-devel \
   openssl-devel \
-  valgrind \
-  ninja-build
+  valgrind
 
 ENV PATH=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/bin:$PATH
 ENV LD_LIBRARY_PATH=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/lib64:/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/lib:$LD_LIBRARY_PATH
@@ -106,6 +103,9 @@ CMD ["/bin/bash"]
 # install test dependencies:
 # - grpcio requires system openssl, bundled crypto fails to build
 RUN dnf install -y \
+  protobuf-devel \
+  protobuf-c-devel \
+  protobuf-lite-devel \
   hdf5-devel \
   python3-h5py \
   git
@@ -129,9 +129,6 @@ RUN pip3 install flatbuffers && \
   git clone https://github.com/microsoft/onnxruntime && \
   cd onnxruntime && git checkout v1.21.0 && \
   git submodule update --init --recursive && \
-  ./build.sh --config Release --parallel 0 --enable_pybind \
-  --build_wheel --enable_training --enable_training_apis \
-  --enable_training_ops --skip_tests --allow_running_as_root \
-  --compile_no_warning_as_error && \
+  ./build.sh --config Release --parallel 0 --enable_pybind --build_wheel --enable_training --enable_training_apis --enable_training_ops --skip_tests --allow_running_as_root && \
   pip3 install ./build/Linux/Release/dist/onnxruntime_training-*.whl && \
   cd .. && /bin/rm -rf ./onnxruntime

.ci/docker/manywheel/build.sh

@@ -27,7 +27,6 @@ fi
 
 MANY_LINUX_VERSION=${MANY_LINUX_VERSION:-}
 DOCKERFILE_SUFFIX=${DOCKERFILE_SUFFIX:-}
-OPENBLAS_VERSION=${OPENBLAS_VERSION:-}
 
 case ${image} in
     manylinux2_28-builder:cpu)
@@ -41,7 +40,6 @@ case ${image} in
         GPU_IMAGE=arm64v8/almalinux:8
         DOCKER_GPU_BUILD_ARG=" --build-arg DEVTOOLSET_VERSION=13 --build-arg NINJA_VERSION=1.12.1"
         MANY_LINUX_VERSION="2_28_aarch64"
-        OPENBLAS_VERSION="v0.3.29"
         ;;
     manylinuxcxx11-abi-builder:cpu-cxx11-abi)
         TARGET=final
@@ -111,7 +109,6 @@ tmp_tag=$(basename "$(mktemp -u)" | tr '[:upper:]' '[:lower:]')
 DOCKER_BUILDKIT=1 docker build  \
     ${DOCKER_GPU_BUILD_ARG} \
     --build-arg "GPU_IMAGE=${GPU_IMAGE}" \
-    --build-arg "OPENBLAS_VERSION=${OPENBLAS_VERSION}" \
     --target "${TARGET}" \
     -t "${tmp_tag}" \
     $@ \

.ci/docker/requirements-ci.txt

@@ -41,11 +41,14 @@ fbscribelogger==0.1.7
 #Pinned versions: 0.1.6
 #test that import:
 
-flatbuffers==24.12.23
+flatbuffers==2.0 ; platform_machine != "s390x"
 #Description: cross platform serialization library
-#Pinned versions: 24.12.23
+#Pinned versions: 2.0
 #test that import:
 
+flatbuffers ; platform_machine == "s390x"
+#Description: cross platform serialization library; Newer version is required on s390x for new python version
+
 hypothesis==5.35.1
 # Pin hypothesis to avoid flakiness: https://github.com/pytorch/pytorch/issues/31136
 #Description: advanced library for generating parametrized tests
@@ -90,7 +93,7 @@ librosa>=0.6.2 ; python_version < "3.11"
 #Pinned versions:
 #test that import:
 
-mypy==1.15.0
+mypy==1.14.0
 # Pin MyPy version because new errors are likely to appear with each release
 #Description: linter
 #Pinned versions: 1.14.0
@@ -163,10 +166,10 @@ pillow==11.0.0
 #Pinned versions: 10.3.0
 #test that import:
 
-protobuf==5.29.4
-#Description:  Google's data interchange format
-#Pinned versions: 5.29.4
-#test that import: test_tensorboard.py, test/onnx/*
+protobuf==3.20.2
+#Description:  Google’s data interchange format
+#Pinned versions: 3.20.1
+#test that import: test_tensorboard.py
 
 psutil
 #Description: information on running processes and system utilization
@@ -334,12 +337,12 @@ sympy==1.13.3
 #Pinned versions:
 #test that import:
 
-onnx==1.18.0
-#Description: Required by onnx tests, and mypy and test_public_bindings.py when checking torch.onnx._internal
+onnx==1.17.0
+#Description: Required by mypy and test_public_bindings.py when checking torch.onnx._internal
 #Pinned versions:
 #test that import:
 
-onnxscript==0.2.6
+onnxscript==0.2.2
 #Description: Required by mypy and test_public_bindings.py when checking torch.onnx._internal
 #Pinned versions:
 #test that import:

.ci/docker/requirements-docs.txt

@@ -15,10 +15,6 @@ sphinxext-opengraph==0.9.1
 #Description: This is used to generate PyTorch docs
 #Pinned versions: 0.9.1
 
-sphinx_sitemap==2.6.0
-#Description: This is used to generate sitemap for PyTorch docs
-#Pinned versions: 2.6.0
-
 matplotlib==3.5.3
 #Description: This is used to generate PyTorch docs
 #Pinned versions: 3.5.3

.ci/docker/triton_version.txt

@@ -1 +1 @@
-3.3.1
+3.3.0

.ci/docker/ubuntu-cuda/Dockerfile

@@ -0,0 +1,170 @@
+ARG UBUNTU_VERSION
+ARG CUDA_VERSION
+ARG IMAGE_NAME
+
+FROM ${IMAGE_NAME} as base
+
+ARG UBUNTU_VERSION
+ARG CUDA_VERSION
+
+ENV DEBIAN_FRONTEND noninteractive
+
+# Install common dependencies (so that this step can be cached separately)
+COPY ./common/install_base.sh install_base.sh
+RUN bash ./install_base.sh && rm install_base.sh
+
+# Install user
+COPY ./common/install_user.sh install_user.sh
+RUN bash ./install_user.sh && rm install_user.sh
+
+# Install katex
+ARG KATEX
+COPY ./common/install_docs_reqs.sh install_docs_reqs.sh
+RUN bash ./install_docs_reqs.sh && rm install_docs_reqs.sh
+
+# Install conda and other packages (e.g., numpy, pytest)
+ARG ANACONDA_PYTHON_VERSION
+ENV ANACONDA_PYTHON_VERSION=$ANACONDA_PYTHON_VERSION
+ENV PATH /opt/conda/envs/py_$ANACONDA_PYTHON_VERSION/bin:/opt/conda/bin:$PATH
+COPY requirements-ci.txt /opt/conda/requirements-ci.txt
+COPY ./common/install_conda.sh install_conda.sh
+COPY ./common/common_utils.sh common_utils.sh
+COPY ./common/install_magma_conda.sh install_magma_conda.sh
+RUN bash ./install_conda.sh && rm install_conda.sh install_magma_conda.sh common_utils.sh /opt/conda/requirements-ci.txt
+
+# Install gcc
+ARG GCC_VERSION
+COPY ./common/install_gcc.sh install_gcc.sh
+RUN bash ./install_gcc.sh && rm install_gcc.sh
+
+# Install clang
+ARG CLANG_VERSION
+COPY ./common/install_clang.sh install_clang.sh
+RUN bash ./install_clang.sh && rm install_clang.sh
+
+# (optional) Install vision packages like OpenCV
+ARG VISION
+COPY ./common/install_vision.sh ./common/cache_vision_models.sh ./common/common_utils.sh ./
+RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
+RUN rm install_vision.sh cache_vision_models.sh common_utils.sh
+ENV INSTALLED_VISION ${VISION}
+
+# (optional) Install UCC
+ARG UCX_COMMIT
+ARG UCC_COMMIT
+ENV UCX_COMMIT $UCX_COMMIT
+ENV UCC_COMMIT $UCC_COMMIT
+ENV UCX_HOME /usr
+ENV UCC_HOME /usr
+ADD ./common/install_ucc.sh install_ucc.sh
+RUN if [ -n "${UCX_COMMIT}" ] && [ -n "${UCC_COMMIT}" ]; then bash ./install_ucc.sh; fi
+RUN rm install_ucc.sh
+
+COPY ./common/install_openssl.sh install_openssl.sh
+ENV OPENSSL_ROOT_DIR /opt/openssl
+RUN bash ./install_openssl.sh
+ENV OPENSSL_DIR /opt/openssl
+
+ARG INDUCTOR_BENCHMARKS
+ARG ANACONDA_PYTHON_VERSION
+ENV ANACONDA_PYTHON_VERSION=$ANACONDA_PYTHON_VERSION
+COPY ./common/install_inductor_benchmark_deps.sh install_inductor_benchmark_deps.sh
+COPY ./common/common_utils.sh common_utils.sh
+COPY ci_commit_pins/huggingface.txt huggingface.txt
+COPY ci_commit_pins/timm.txt timm.txt
+RUN if [ -n "${INDUCTOR_BENCHMARKS}" ]; then bash ./install_inductor_benchmark_deps.sh; fi
+RUN rm install_inductor_benchmark_deps.sh common_utils.sh timm.txt huggingface.txt
+
+ARG TRITON
+
+FROM base as triton-builder
+# Install triton, this needs to be done before sccache because the latter will
+# try to reach out to S3, which docker build runners don't have access
+COPY ./common/install_triton.sh install_triton.sh
+COPY ./common/common_utils.sh common_utils.sh
+COPY ci_commit_pins/triton.txt triton.txt
+COPY triton_version.txt triton_version.txt
+RUN bash ./install_triton.sh
+
+FROM base as final
+COPY --from=triton-builder /opt/triton /opt/triton
+RUN if [ -n "${TRITON}" ]; then pip install /opt/triton/*.whl; chown -R jenkins:jenkins /opt/conda; fi
+RUN rm -rf /opt/triton
+
+ARG HALIDE
+# Build and install halide
+COPY ./common/install_halide.sh install_halide.sh
+COPY ./common/common_utils.sh common_utils.sh
+COPY ci_commit_pins/halide.txt halide.txt
+RUN if [ -n "${HALIDE}" ]; then bash ./install_halide.sh; fi
+RUN rm install_halide.sh common_utils.sh halide.txt
+
+# Install ccache/sccache (do this last, so we get priority in PATH)
+COPY ./common/install_cache.sh install_cache.sh
+ENV PATH /opt/cache/bin:$PATH
+# See https://github.com/pytorch/pytorch/issues/82174
+# TODO(sdym@fb.com):
+# check if this is needed after full off Xenial migration
+ENV CARGO_NET_GIT_FETCH_WITH_CLI true
+RUN bash ./install_cache.sh && rm install_cache.sh
+ENV CMAKE_CUDA_COMPILER_LAUNCHER=/opt/cache/bin/sccache
+
+# Add jni.h for java host build
+COPY ./common/install_jni.sh install_jni.sh
+COPY ./java/jni.h jni.h
+RUN bash ./install_jni.sh && rm install_jni.sh
+
+# Install Open MPI for CUDA
+COPY ./common/install_openmpi.sh install_openmpi.sh
+RUN if [ -n "${CUDA_VERSION}" ]; then bash install_openmpi.sh; fi
+RUN rm install_openmpi.sh
+
+# Include BUILD_ENVIRONMENT environment variable in image
+ARG BUILD_ENVIRONMENT
+ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
+
+# AWS specific CUDA build guidance
+ENV TORCH_CUDA_ARCH_LIST Maxwell
+ENV TORCH_NVCC_FLAGS "-Xfatbin -compress-all"
+ENV CUDA_PATH /usr/local/cuda
+
+# Install LLVM dev version (Defined in the pytorch/builder github repository)
+COPY --from=pytorch/llvm:9.0.1 /opt/llvm /opt/llvm
+
+# Install CUDNN
+ARG CUDNN_VERSION
+ARG CUDA_VERSION
+COPY ./common/install_cudnn.sh install_cudnn.sh
+RUN if [ -n "${CUDNN_VERSION}" ]; then bash install_cudnn.sh; fi
+RUN rm install_cudnn.sh
+
+# Install CUSPARSELT
+ARG CUDA_VERSION
+COPY ./common/install_cusparselt.sh install_cusparselt.sh
+RUN bash install_cusparselt.sh
+RUN rm install_cusparselt.sh
+
+# Install NCCL
+ARG CUDA_VERSION
+COPY ./common/install_nccl.sh install_nccl.sh
+COPY ./ci_commit_pins/nccl-cu* /ci_commit_pins/
+RUN bash install_nccl.sh
+RUN rm install_nccl.sh /ci_commit_pins/nccl-cu*
+ENV USE_SYSTEM_NCCL=1
+ENV NCCL_INCLUDE_DIR="/usr/local/cuda/include/"
+ENV NCCL_LIB_DIR="/usr/local/cuda/lib64/"
+
+# Install CUDSS
+ARG CUDA_VERSION
+COPY ./common/install_cudss.sh install_cudss.sh
+RUN bash install_cudss.sh
+RUN rm install_cudss.sh
+
+# Delete /usr/local/cuda-11.X/cuda-11.X symlinks
+RUN if [ -h /usr/local/cuda-11.6/cuda-11.6 ]; then rm /usr/local/cuda-11.6/cuda-11.6; fi
+RUN if [ -h /usr/local/cuda-11.7/cuda-11.7 ]; then rm /usr/local/cuda-11.7/cuda-11.7; fi
+RUN if [ -h /usr/local/cuda-12.1/cuda-12.1 ]; then rm /usr/local/cuda-12.1/cuda-12.1; fi
+RUN if [ -h /usr/local/cuda-12.4/cuda-12.4 ]; then rm /usr/local/cuda-12.4/cuda-12.4; fi
+
+USER jenkins
+CMD ["bash"]

.ci/manywheel/build_common.sh

@@ -18,10 +18,12 @@ retry () {
     $*  || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
 }
 
-PLATFORM=""
+PLATFORM="manylinux2014_x86_64"
 # TODO move this into the Docker images
 OS_NAME=$(awk -F= '/^NAME/{print $2}' /etc/os-release)
-if [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
+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
     PLATFORM="manylinux_2_28_x86_64"
 elif [[ "$OS_NAME" == *"Red Hat Enterprise Linux"* ]]; then
@@ -34,9 +36,6 @@ elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
 
     retry apt-get update
     retry apt-get -y install zip openssl
-else
-    echo "Unknown OS: '$OS_NAME'"
-    exit 1
 fi
 
 # We use the package name to test the package by passing this to 'pip install'
@@ -80,6 +79,8 @@ if [[ -e /opt/openssl ]]; then
     export CMAKE_INCLUDE_PATH="/opt/openssl/include":$CMAKE_INCLUDE_PATH
 fi
 
+
+
 mkdir -p /tmp/$WHEELHOUSE_DIR
 
 export PATCHELF_BIN=/usr/local/bin/patchelf

.ci/manywheel/build_cuda.sh

@@ -15,9 +15,6 @@ export INSTALL_TEST=0 # dont install test binaries into site-packages
 export USE_CUPTI_SO=0
 export USE_CUSPARSELT=${USE_CUSPARSELT:-1} # Enable if not disabled by libtorch build
 export USE_CUFILE=${USE_CUFILE:-1}
-export USE_SYSTEM_NCCL=1
-export NCCL_INCLUDE_DIR="/usr/local/cuda/include/"
-export NCCL_LIB_DIR="/usr/local/cuda/lib64/"
 
 # Keep an array of cmake variables to add to
 if [[ -z "$CMAKE_ARGS" ]]; then
@@ -39,8 +36,10 @@ if [[ -n "$DESIRED_CUDA" ]]; then
     if [[ ${DESIRED_CUDA} =~ ^[0-9]+\.[0-9]+$ ]]; then
         CUDA_VERSION=${DESIRED_CUDA}
     else
-        # cu126, cu128 etc...
-        if [[ ${#DESIRED_CUDA} -eq 5 ]]; then
+        # 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
@@ -62,6 +61,10 @@ case ${CUDA_VERSION} in
         TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};9.0"
         EXTRA_CAFFE2_CMAKE_FLAGS+=("-DATEN_NO_TEST=ON")
         ;;
+    12.4)
+        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")
@@ -88,15 +91,14 @@ fi
 mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR" || true
 
 OS_NAME=$(awk -F= '/^NAME/{print $2}' /etc/os-release)
-if [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
+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"
-else
-    echo "Unknown OS: '$OS_NAME'"
-    exit 1
 fi
 
 DEPS_LIST=(
@@ -106,8 +108,26 @@ DEPS_SONAME=(
     "libgomp.so.1"
 )
 
+# CUDA 11.8 have to ship the libcusparseLt.so.0 with the binary
+# since nvidia-cusparselt-cu11 is not available in PYPI
+if [[ $USE_CUSPARSELT == "1" && $CUDA_VERSION == "11.8" ]]; then
+        DEPS_SONAME+=(
+            "libcusparseLt.so.0"
+        )
+        DEPS_LIST+=(
+            "/usr/local/cuda/lib64/libcusparseLt.so.0"
+        )
+fi
 
-# CUDA_VERSION 12.6, 12.8
+
+# Turn USE_CUFILE off for CUDA 11.8, 12.4 since nvidia-cufile-cu11 and 1.9.0.20 are
+# not available in PYPI
+if [[ $CUDA_VERSION == "11.8" || $CUDA_VERSION == "12.4" ]]; then
+    export USE_CUFILE=0
+fi
+
+
+# CUDA_VERSION 12.4, 12.6, 12.8
 if [[ $CUDA_VERSION == 12* ]]; then
     export USE_STATIC_CUDNN=0
     # Try parallelizing nvcc as well
@@ -131,8 +151,6 @@ if [[ $CUDA_VERSION == 12* ]]; then
             "/usr/local/cuda/lib64/libnvToolsExt.so.1"
             "/usr/local/cuda/lib64/libnvrtc.so.12"
             "/usr/local/cuda/lib64/libnvrtc-builtins.so"
-            "/usr/local/cuda/lib64/libcufile.so.0"
-            "/usr/local/cuda/lib64/libcufile_rdma.so.1"
         )
         DEPS_SONAME+=(
             "libcudnn_adv.so.9"
@@ -150,9 +168,17 @@ if [[ $CUDA_VERSION == 12* ]]; then
             "libnvToolsExt.so.1"
             "libnvrtc.so.12"
             "libnvrtc-builtins.so"
-            "libcufile.so.0"
-            "libcufile_rdma.so.1"
         )
+        if [[ $USE_CUFILE == 1 ]]; then
+            DEPS_LIST+=(
+                "/usr/local/cuda/lib64/libcufile.so.0"
+                "/usr/local/cuda/lib64/libcufile_rdma.so.1"
+            )
+            DEPS_SONAME+=(
+                "libcufile.so.0"
+                "libcufile_rdma.so.1"
+            )
+        fi
     else
         echo "Using nvidia libs from pypi."
         CUDA_RPATHS=(
@@ -165,43 +191,34 @@ if [[ $CUDA_VERSION == 12* ]]; then
             '$ORIGIN/../../nvidia/curand/lib'
             '$ORIGIN/../../nvidia/cusolver/lib'
             '$ORIGIN/../../nvidia/cusparse/lib'
-            '$ORIGIN/../../nvidia/cusparselt/lib'
             '$ORIGIN/../../cusparselt/lib'
             '$ORIGIN/../../nvidia/nccl/lib'
-            '$ORIGIN/../../nvidia/nvshmem/lib'
             '$ORIGIN/../../nvidia/nvtx/lib'
-            '$ORIGIN/../../nvidia/cufile/lib'
         )
+        if [[ $USE_CUFILE == 1 ]]; then
+            CUDA_RPATHS+=(
+                '$ORIGIN/../../nvidia/cufile/lib'
+            )
+        fi
         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
-    # Turn USE_CUFILE off for CUDA 11.8 since nvidia-cufile-cu11 and 1.9.0.20 are
-    # not available in PYPI
-    export USE_CUFILE=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
 
-    # CUDA 11.8 have to ship the libcusparseLt.so.0 with the binary
-    # since nvidia-cusparselt-cu11 is not available in PYPI
-    if [[ $USE_CUSPARSELT == "1" ]]; then
-        DEPS_SONAME+=(
-            "libcusparseLt.so.0"
-        )
-        DEPS_LIST+=(
-            "/usr/local/cuda/lib64/libcusparseLt.so.0"
-        )
-    fi
-
     if [[ -z "$PYTORCH_EXTRA_INSTALL_REQUIREMENTS" ]]; then
         echo "Bundling with cudnn and cublas."
         DEPS_LIST+=(
@@ -256,9 +273,12 @@ elif [[ $CUDA_VERSION == "11.8" ]]; then
         export LIB_SO_RPATH=$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"

.ci/manywheel/build_libtorch.sh

@@ -22,7 +22,9 @@ retry () {
 
 # TODO move this into the Docker images
 OS_NAME=`awk -F= '/^NAME/{print $2}' /etc/os-release`
-if [[ "$OS_NAME" == *"AlmaLinux"* ]]; then
+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 [[ "$OS_NAME" == *"Red Hat Enterprise Linux"* ]]; then
     retry dnf install -q -y zip openssl
@@ -33,9 +35,6 @@ elif [[ "$OS_NAME" == *"Ubuntu"* ]]; then
     sed -i 's/.*nvidia.*/# &/' $(find /etc/apt/ -type f -name "*.list")
     retry apt-get update
     retry apt-get -y install zip openssl
-else
-    echo "Unknown OS: '$OS_NAME'"
-    exit 1
 fi
 
 # Version: setup.py uses $PYTORCH_BUILD_VERSION.post$PYTORCH_BUILD_NUMBER if

.ci/manywheel/build_rocm.sh

@@ -95,7 +95,6 @@ ROCM_SO_FILES=(
     "libroctracer64.so"
     "libroctx64.so"
     "libhipblaslt.so"
-    "libhipsparselt.so"
     "libhiprtc.so"
 )
 
@@ -187,28 +186,20 @@ do
     OS_SO_FILES[${#OS_SO_FILES[@]}]=$file_name # Append lib to array
 done
 
-ARCH=$(echo $PYTORCH_ROCM_ARCH | sed 's/;/|/g') # Replace ; seperated arch list to bar for grep
-
 # rocBLAS library files
 ROCBLAS_LIB_SRC=$ROCM_HOME/lib/rocblas/library
 ROCBLAS_LIB_DST=lib/rocblas/library
-ROCBLAS_ARCH_SPECIFIC_FILES=$(ls $ROCBLAS_LIB_SRC | grep -E $ARCH)
-ROCBLAS_OTHER_FILES=$(ls $ROCBLAS_LIB_SRC | grep -v gfx)
-ROCBLAS_LIB_FILES=($ROCBLAS_ARCH_SPECIFIC_FILES $OTHER_FILES)
+ARCH=$(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=$ROCM_HOME/lib/hipblaslt/library
 HIPBLASLT_LIB_DST=lib/hipblaslt/library
-HIPBLASLT_ARCH_SPECIFIC_FILES=$(ls $HIPBLASLT_LIB_SRC | grep -E $ARCH)
-HIPBLASLT_OTHER_FILES=$(ls $HIPBLASLT_LIB_SRC | grep -v gfx)
-HIPBLASLT_LIB_FILES=($HIPBLASLT_ARCH_SPECIFIC_FILES $HIPBLASLT_OTHER_FILES)
-
-# hipsparselt library files
-HIPSPARSELT_LIB_SRC=$ROCM_HOME/lib/hipsparselt/library
-HIPSPARSELT_LIB_DST=lib/hipsparselt/library
-HIPSPARSELT_ARCH_SPECIFIC_FILES=$(ls $HIPSPARSELT_LIB_SRC | grep -E $ARCH)
-#HIPSPARSELT_OTHER_FILES=$(ls $HIPSPARSELT_LIB_SRC | grep -v gfx)
-HIPSPARSELT_LIB_FILES=($HIPSPARSELT_ARCH_SPECIFIC_FILES $HIPSPARSELT_OTHER_FILES)
+ARCH_SPECIFIC_FILES=$(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=()
@@ -243,14 +234,12 @@ DEPS_SONAME=(
 DEPS_AUX_SRCLIST=(
     "${ROCBLAS_LIB_FILES[@]/#/$ROCBLAS_LIB_SRC/}"
     "${HIPBLASLT_LIB_FILES[@]/#/$HIPBLASLT_LIB_SRC/}"
-    "${HIPSPARSELT_LIB_FILES[@]/#/$HIPSPARSELT_LIB_SRC/}"
     "/opt/amdgpu/share/libdrm/amdgpu.ids"
 )
 
 DEPS_AUX_DSTLIST=(
     "${ROCBLAS_LIB_FILES[@]/#/$ROCBLAS_LIB_DST/}"
     "${HIPBLASLT_LIB_FILES[@]/#/$HIPBLASLT_LIB_DST/}"
-    "${HIPSPARSELT_LIB_FILES[@]/#/$HIPSPARSELT_LIB_DST/}"
     "share/libdrm/amdgpu.ids"
 )
 

.ci/pytorch/build.sh

@@ -52,6 +52,12 @@ fi
 export USE_LLVM=/opt/llvm
 export LLVM_DIR=/opt/llvm/lib/cmake/llvm
 
+if [[ "$BUILD_ENVIRONMENT" == *executorch* ]]; then
+  # To build test_edge_op_registration
+  export BUILD_EXECUTORCH=ON
+  export USE_CUDA=0
+fi
+
 if ! which conda; then
   # In ROCm CIs, we are doing cross compilation on build machines with
   # intel cpu and later run tests on machines with amd cpu.

.ci/pytorch/macos-build.sh

@@ -40,7 +40,7 @@ if [[ ${BUILD_ENVIRONMENT} == *"distributed"* ]]; then
 else
   # Explicitly set USE_DISTRIBUTED=0 to align with the default build config on mac. This also serves as the sole CI config that tests
   # that building with USE_DISTRIBUTED=0 works at all. See https://github.com/pytorch/pytorch/issues/86448
-  USE_DISTRIBUTED=0 USE_OPENMP=1 MACOSX_DEPLOYMENT_TARGET=11.0 WERROR=1 BUILD_TEST=OFF USE_PYTORCH_METAL=1 python setup.py bdist_wheel --plat-name macosx_11_0_arm64
+  USE_DISTRIBUTED=0 USE_OPENMP=1 MACOSX_DEPLOYMENT_TARGET=11.0 WERROR=1 BUILD_TEST=OFF USE_PYTORCH_METAL=1 python setup.py bdist_wheel
 fi
 if which sccache > /dev/null; then
   print_sccache_stats

.ci/pytorch/macos-common.sh

@@ -20,4 +20,14 @@ print_cmake_info() {
   CONDA_INSTALLATION_DIR=$(dirname "$CMAKE_EXEC")
   # Print all libraries under cmake rpath for debugging
   ls -la "$CONDA_INSTALLATION_DIR/../lib"
+
+  export CMAKE_EXEC
+  # Explicitly add conda env lib folder to cmake rpath to address the flaky issue
+  # where cmake dependencies couldn't be found. This seems to point to how conda
+  # links $CMAKE_EXEC to its package cache when cloning a new environment
+  install_name_tool -add_rpath @executable_path/../lib "${CMAKE_EXEC}" || true
+  # Adding the rpath will invalidate cmake signature, so signing it again here
+  # to trust the executable. EXC_BAD_ACCESS (SIGKILL (Code Signature Invalid))
+  # with an exit code 137 otherwise
+  codesign -f -s - "${CMAKE_EXEC}" || true
 }

.ci/pytorch/macos-test.sh

@@ -165,7 +165,6 @@ test_jit_hooks() {
 torchbench_setup_macos() {
   git clone --recursive https://github.com/pytorch/vision torchvision
   git clone --recursive https://github.com/pytorch/audio torchaudio
-  brew install jpeg-turbo libpng
 
   pushd torchvision
   git fetch
@@ -180,8 +179,7 @@ torchbench_setup_macos() {
   git checkout "$(cat ../.github/ci_commit_pins/audio.txt)"
   git submodule update --init --recursive
   python setup.py clean
-  #TODO: Remove me, when figure out how to make TorchAudio find brew installed openmp
-  USE_OPENMP=0 python setup.py develop
+  python setup.py develop
   popd
 
   # Shellcheck doesn't like it when you pass no arguments to a function that can take args. See https://www.shellcheck.net/wiki/SC2120
@@ -189,8 +187,9 @@ torchbench_setup_macos() {
   checkout_install_torchbench
 }
 
-pip_benchmark_deps() {
-  python -mpip install --no-input astunparse requests cython scikit-learn
+conda_benchmark_deps() {
+  conda install -y astunparse numpy scipy ninja pyyaml setuptools cmake typing-extensions requests protobuf numba cython scikit-learn
+  conda install -y -c conda-forge librosa
 }
 
 
@@ -198,7 +197,7 @@ test_torchbench_perf() {
   print_cmake_info
 
   echo "Launching torchbench setup"
-  pip_benchmark_deps
+  conda_benchmark_deps
   torchbench_setup_macos
 
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
@@ -225,7 +224,7 @@ test_torchbench_smoketest() {
   print_cmake_info
 
   echo "Launching torchbench setup"
-  pip_benchmark_deps
+  conda_benchmark_deps
   # shellcheck disable=SC2119,SC2120
   torchbench_setup_macos
 
@@ -233,8 +232,8 @@ test_torchbench_smoketest() {
   mkdir -p "$TEST_REPORTS_DIR"
 
   local device=mps
-  local models=(hf_T5 llama BERT_pytorch dcgan hf_GPT2 yolov3 resnet152 sam pytorch_unet stable_diffusion_text_encoder speech_transformer Super_SloMo doctr_det_predictor doctr_reco_predictor)
-  local hf_models=(GoogleFnet YituTechConvBert Speech2Text2ForCausalLM)
+  local models=(hf_T5 llama BERT_pytorch dcgan hf_GPT2 yolov3 resnet152 sam pytorch_unet stable_diffusion_text_encoder speech_transformer Super_SloMo)
+  local hf_models=(GoogleFnet YituTechConvBert)
 
   for backend in eager inductor; do
 
@@ -259,10 +258,10 @@ test_torchbench_smoketest() {
         if [ "$backend" == "inductor" ]; then
           PYTHONPATH="$(pwd)"/torchbench python benchmarks/dynamo/huggingface.py \
             --performance --only "$model" --backend "$backend" --inference --devices "$device" "$dtype_arg" \
-            --output "$TEST_REPORTS_DIR/inductor_${backend}_huggingface_${dtype}_inference_${device}_performance.csv" || true
+            --output "$TEST_REPORTS_DIR/inductor_${backend}_torchbench_${dtype}_inference_${device}_performance.csv" || true
           PYTHONPATH="$(pwd)"/torchbench python benchmarks/dynamo/huggingface.py \
             --accuracy --only "$model" --backend "$backend" --inference --devices "$device" "$dtype_arg" \
-            --output "$TEST_REPORTS_DIR/inductor_${backend}_huggingface_${dtype}_inference_${device}_accuracy.csv" || true
+            --output "$TEST_REPORTS_DIR/inductor_${backend}_torchbench_${dtype}_inference_${device}_accuracy.csv" || true
         fi
       done
     done
@@ -290,7 +289,7 @@ test_hf_perf() {
   print_cmake_info
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
   mkdir -p "$TEST_REPORTS_DIR"
-  pip_benchmark_deps
+  conda_benchmark_deps
   torchbench_setup_macos
 
   echo "Launching HuggingFace training perf run"
@@ -306,7 +305,7 @@ test_timm_perf() {
   print_cmake_info
   TEST_REPORTS_DIR=$(pwd)/test/test-reports
   mkdir -p "$TEST_REPORTS_DIR"
-  pip_benchmark_deps
+  conda_benchmark_deps
   torchbench_setup_macos
 
   echo "Launching timm training perf run"

.ci/pytorch/smoke_test/check_gomp.py

@@ -46,9 +46,6 @@ def get_gomp_thread():
 
     # use the default gomp path of AlmaLinux OS
     libgomp_path = "/usr/lib64/libgomp.so.1"
-    # if it does not exist, try Ubuntu path
-    if not os.path.exists(libgomp_path):
-        libgomp_path = f"/usr/lib/{os.uname().machine}-linux-gnu/libgomp.so.1"
 
     os.environ["GOMP_CPU_AFFINITY"] = "0-3"
 

.ci/pytorch/test.sh

@@ -324,12 +324,6 @@ test_python_smoke() {
   assert_git_not_dirty
 }
 
-test_h100_distributed() {
-  # Distributed tests at H100
-  time python test/run_test.py --include distributed/_composable/test_composability/test_pp_composability.py  $PYTHON_TEST_EXTRA_OPTION --upload-artifacts-while-running
-  assert_git_not_dirty
-}
-
 test_lazy_tensor_meta_reference_disabled() {
   export TORCH_DISABLE_FUNCTIONALIZATION_META_REFERENCE=1
   echo "Testing lazy tensor operations without meta reference"
@@ -826,7 +820,16 @@ test_inductor_torchbench_smoketest_perf() {
   done
 }
 
+test_inductor_get_core_number() {
+  if [[ "${TEST_CONFIG}" == *aarch64* ]]; then
+    echo "$(($(lscpu | grep 'Cluster(s):' | awk '{print $2}') * $(lscpu | grep 'Core(s) per cluster:' | awk '{print $4}')))"
+  else
+    echo "$(($(lscpu | grep 'Socket(s):' | awk '{print $2}') * $(lscpu | grep 'Core(s) per socket:' | awk '{print $4}')))"
+  fi
+}
+
 test_inductor_set_cpu_affinity(){
+  #set jemalloc
   JEMALLOC_LIB="$(find /usr/lib -name libjemalloc.so.2)"
   export LD_PRELOAD="$JEMALLOC_LIB":"$LD_PRELOAD"
   export MALLOC_CONF="oversize_threshold:1,background_thread:true,metadata_thp:auto,dirty_decay_ms:-1,muzzy_decay_ms:-1"
@@ -838,23 +841,14 @@ test_inductor_set_cpu_affinity(){
     export KMP_AFFINITY=granularity=fine,compact,1,0
     export KMP_BLOCKTIME=1
   fi
-
-  # Use nproc here instead of lscpu because it takes into account cgroups slice
-  cpus=$(nproc)
-  thread_per_core=$(lscpu | grep 'Thread(s) per core:' | awk '{print $4}')
-  cores=$((cpus / thread_per_core))
-
-  # Set number of cores to 16 on aarch64 for performance runs
+  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
-
-  # Handle cgroups slice start and end CPU
-  start_cpu=$(python -c 'import os; print(min(os.sched_getaffinity(0)))')
-  # Leaving one physical CPU for other tasks
-  end_cpu=$(($(python -c 'import os; print(max(os.sched_getaffinity(0)))') - thread_per_core))
-  export TASKSET="taskset -c $start_cpu-$end_cpu"
+  end_core=$((cores-1))
+  export TASKSET="taskset -c 0-$end_core"
 }
 
 test_inductor_torchbench_cpu_smoketest_perf(){
@@ -1526,7 +1520,7 @@ test_executorch() {
 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 test_autograd test_binary_ufuncs test_complex test_spectral_ops \
-        test_foreach test_reductions test_unary_ufuncs test_tensor_creation_ops test_ops test_cpp_extensions_open_device_registration \
+        test_foreach test_reductions test_unary_ufuncs test_tensor_creation_ops test_ops \
         --shard "$SHARD_NUMBER" "$NUM_TEST_SHARDS" --verbose
 
   # Dynamo tests
@@ -1560,8 +1554,7 @@ test_operator_benchmark() {
 
   cd "${TEST_DIR}"/benchmarks/operator_benchmark
   $TASKSET python -m benchmark_all_test --device "$1" --tag-filter "$2" \
-      --output-csv "${TEST_REPORTS_DIR}/operator_benchmark_eager_float32_cpu.csv" \
-      --output-json-for-dashboard "${TEST_REPORTS_DIR}/operator_benchmark_eager_float32_cpu.json" \
+      --output-dir "${TEST_REPORTS_DIR}/operator_benchmark_eager_float32_cpu.csv"
 
   pip_install pandas
   python check_perf_csv.py \
@@ -1646,7 +1639,7 @@ elif [[ "${TEST_CONFIG}" == *torchbench* ]]; then
     install_torchaudio cuda
   fi
   install_torchvision
-  TORCH_CUDA_ARCH_LIST="8.0;8.6" install_torchao
+  TORCH_CUDA_ARCH_LIST="8.0;8.6" pip_install git+https://github.com/pytorch/ao.git
   id=$((SHARD_NUMBER-1))
   # https://github.com/opencv/opencv-python/issues/885
   pip_install opencv-python==4.8.0.74
@@ -1731,8 +1724,6 @@ elif [[ "${BUILD_ENVIRONMENT}" == *xpu* ]]; then
   test_xpu_bin
 elif [[ "${TEST_CONFIG}" == smoke ]]; then
   test_python_smoke
-elif [[ "${TEST_CONFIG}" == h100_distributed ]]; then
-  test_h100_distributed
 else
   install_torchvision
   install_monkeytype

.ci/pytorch/win-test.sh

@@ -38,7 +38,7 @@ if [[ "$BUILD_ENVIRONMENT" == *cuda* ]]; then
 fi
 
 # TODO: Move both of them to Windows AMI
-python -m pip install pytest-rerunfailures==10.3 pytest-cpp==2.3.0 tensorboard==2.13.0 protobuf==5.29.4 pytest-subtests==0.13.1
+python -m pip install pytest-rerunfailures==10.3 pytest-cpp==2.3.0 tensorboard==2.13.0 pytest-subtests==0.13.1
 
 # Install Z3 optional dependency for Windows builds.
 python -m pip install z3-solver==4.12.2.0

.ci/pytorch/windows/arm64/bootstrap_libuv.bat

@@ -7,7 +7,7 @@ if not exist "%DOWNLOADS_DIR%" mkdir %DOWNLOADS_DIR%
 if not exist "%DEPENDENCIES_DIR%" mkdir %DEPENDENCIES_DIR%
 
 :: activate visual studio
-call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" arm64
+call "%DEPENDENCIES_DIR%\VSBuildTools\VC\Auxiliary\Build\vcvarsall.bat" arm64
 where cl.exe
 
 cd %DEPENDENCIES_DIR%

.ci/pytorch/windows/arm64/bootstrap_openblas.bat

@@ -7,7 +7,7 @@ if not exist "%DOWNLOADS_DIR%" mkdir %DOWNLOADS_DIR%
 if not exist "%DEPENDENCIES_DIR%" mkdir %DEPENDENCIES_DIR%
 
 :: activate visual studio
-call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" arm64
+call "%DEPENDENCIES_DIR%\VSBuildTools\VC\Auxiliary\Build\vcvarsall.bat" arm64
 where cl.exe
 
 :: Clone OpenBLAS

.ci/pytorch/windows/arm64/bootstrap_tests.bat

@@ -2,7 +2,7 @@
 cd %PYTORCH_ROOT%
 
 :: activate visual studio
-call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" arm64
+call "%DEPENDENCIES_DIR%\VSBuildTools\VC\Auxiliary\Build\vcvarsall.bat" arm64
 where cl.exe
 
 :: create virtual environment

.ci/pytorch/windows/arm64/build_libtorch.bat

@@ -21,7 +21,7 @@ if %ENABLE_APL% == 1 (
 )
 
 :: activate visual studio
-call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" arm64
+call "%DEPENDENCIES_DIR%\VSBuildTools\VC\Auxiliary\Build\vcvarsall.bat" arm64
 where cl.exe
 
 :: change to source directory

.ci/pytorch/windows/arm64/build_pytorch.bat

@@ -21,7 +21,7 @@ if %ENABLE_APL% == 1 (
 )
 
 :: activate visual studio
-call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" arm64
+call "%DEPENDENCIES_DIR%\VSBuildTools\VC\Auxiliary\Build\vcvarsall.bat" arm64
 where cl.exe
 
 :: change to source directory

.ci/pytorch/windows/arm64/smoke_test.bat

@@ -33,7 +33,7 @@ pushd tmp
 set VC_VERSION_LOWER=14
 set VC_VERSION_UPPER=36
 
-call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" arm64
+call "%DEPENDENCIES_DIR%\VSBuildTools\VC\Auxiliary\Build\vcvarsall.bat" arm64
 
 set install_root=%CD%
 set INCLUDE=%INCLUDE%;%install_root%\include;%install_root%\include\torch\csrc\api\include

.ci/pytorch/windows/cuda124.bat

@@ -0,0 +1,59 @@
+@echo off
+
+set MODULE_NAME=pytorch
+
+IF NOT EXIST "setup.py" IF NOT EXIST "%MODULE_NAME%" (
+    call internal\clone.bat
+    cd %~dp0
+) ELSE (
+    call internal\clean.bat
+)
+IF ERRORLEVEL 1 goto :eof
+
+call internal\check_deps.bat
+IF ERRORLEVEL 1 goto :eof
+
+REM Check for optional components
+
+set USE_CUDA=
+set CMAKE_GENERATOR=Visual Studio 15 2017 Win64
+
+IF "%NVTOOLSEXT_PATH%"=="" (
+    IF EXIST "C:\Program Files\NVIDIA Corporation\NvToolsExt\lib\x64\nvToolsExt64_1.lib"  (
+        set NVTOOLSEXT_PATH=C:\Program Files\NVIDIA Corporation\NvToolsExt
+    ) ELSE (
+        echo NVTX ^(Visual Studio Extension ^for CUDA^) ^not installed, failing
+        exit /b 1
+    )
+)
+
+IF "%CUDA_PATH_V124%"=="" (
+    IF EXIST "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\bin\nvcc.exe" (
+        set "CUDA_PATH_V124=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4"
+    ) ELSE (
+        echo CUDA 12.4 not found, failing
+        exit /b 1
+    )
+)
+
+IF "%BUILD_VISION%" == "" (
+    set TORCH_CUDA_ARCH_LIST=6.1;7.0;7.5;8.0;8.6;9.0
+    set TORCH_NVCC_FLAGS=-Xfatbin -compress-all
+) ELSE (
+    set NVCC_FLAGS=-D__CUDA_NO_HALF_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_50,code=sm_50 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_70,code=sm_70 -gencode=arch=compute_75,code=sm_75 -gencode=arch=compute_80,code=compute_80 -gencode=arch=compute_86,code=compute_86 -gencode=arch=compute_90,code=compute_90
+)
+
+set "CUDA_PATH=%CUDA_PATH_V124%"
+set "PATH=%CUDA_PATH_V124%\bin;%PATH%"
+
+:optcheck
+
+call internal\check_opts.bat
+IF ERRORLEVEL 1 goto :eof
+
+if exist "%NIGHTLIES_PYTORCH_ROOT%" cd %NIGHTLIES_PYTORCH_ROOT%\..
+call  %~dp0\internal\copy.bat
+IF ERRORLEVEL 1 goto :eof
+
+call  %~dp0\internal\setup.bat
+IF ERRORLEVEL 1 goto :eof

.ci/pytorch/windows/internal/cuda_install.bat

@@ -58,6 +58,33 @@ xcopy /Y "%SRC_DIR%\temp_build\zlib\dll_x64\*.dll" "C:\Windows\System32"
 
 goto cuda_common
 
+:cuda124
+
+set CUDA_INSTALL_EXE=cuda_12.4.0_551.61_windows.exe
+if not exist "%SRC_DIR%\temp_build\%CUDA_INSTALL_EXE%" (
+    curl -k -L "https://ossci-windows.s3.amazonaws.com/%CUDA_INSTALL_EXE%" --output "%SRC_DIR%\temp_build\%CUDA_INSTALL_EXE%" & REM @lint-ignore
+    if errorlevel 1 exit /b 1
+    set "CUDA_SETUP_FILE=%SRC_DIR%\temp_build\%CUDA_INSTALL_EXE%"
+    set "ARGS=cuda_profiler_api_12.4 thrust_12.4 nvcc_12.4 cuobjdump_12.4 nvprune_12.4 nvprof_12.4 cupti_12.4 cublas_12.4 cublas_dev_12.4 cudart_12.4 cufft_12.4 cufft_dev_12.4 curand_12.4 curand_dev_12.4 cusolver_12.4 cusolver_dev_12.4 cusparse_12.4 cusparse_dev_12.4 npp_12.4 npp_dev_12.4 nvrtc_12.4 nvrtc_dev_12.4 nvml_dev_12.4 nvjitlink_12.4 nvtx_12.4"
+)
+
+set CUDNN_FOLDER=cudnn-windows-x86_64-9.5.0.50_cuda12-archive
+set CUDNN_LIB_FOLDER="lib"
+set "CUDNN_INSTALL_ZIP=%CUDNN_FOLDER%.zip"
+if not exist "%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%" (
+    curl -k -L "http://s3.amazonaws.com/ossci-windows/%CUDNN_INSTALL_ZIP%" --output "%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%" & REM @lint-ignore
+    if errorlevel 1 exit /b 1
+    set "CUDNN_SETUP_FILE=%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%"
+)
+
+@REM cuDNN 8.3+ required zlib to be installed on the path
+echo Installing ZLIB dlls
+curl -k -L "http://s3.amazonaws.com/ossci-windows/zlib123dllx64.zip" --output "%SRC_DIR%\temp_build\zlib123dllx64.zip"
+7z x "%SRC_DIR%\temp_build\zlib123dllx64.zip" -o"%SRC_DIR%\temp_build\zlib"
+xcopy /Y "%SRC_DIR%\temp_build\zlib\dll_x64\*.dll" "C:\Windows\System32"
+
+goto cuda_common
+
 :cuda126
 
 set CUDA_INSTALL_EXE=cuda_12.6.2_560.94_windows.exe

.ci/pytorch/windows/internal/vc_install_helper.bat

@@ -3,8 +3,6 @@ if "%VC_YEAR%" == "2022" powershell windows/internal/vs2022_install.ps1
 
 set VC_VERSION_LOWER=17
 set VC_VERSION_UPPER=18
-:: Please don't delete VS2019 as an alternative, in case some Windows compiler issue.
-:: Reference: https://github.com/pytorch/pytorch/issues/145702#issuecomment-2858693930
 if "%VC_YEAR%" == "2019" (
     set VC_VERSION_LOWER=16
     set VC_VERSION_UPPER=17

.circleci/scripts/binary_windows_build.sh

@@ -9,7 +9,7 @@ if [[ "$OS" != "windows-arm64" ]]; then
     export USE_SCCACHE=1
     export SCCACHE_BUCKET=ossci-compiler-cache
     export SCCACHE_IGNORE_SERVER_IO_ERROR=1
-    export VC_YEAR=2022
+    export VC_YEAR=2019
 fi
 
 if [[ "$DESIRED_CUDA" == 'xpu' ]]; then

.circleci/scripts/binary_windows_test.sh

@@ -4,7 +4,7 @@ set -eux -o pipefail
 source "${BINARY_ENV_FILE:-/c/w/env}"
 
 export CUDA_VERSION="${DESIRED_CUDA/cu/}"
-export VC_YEAR=2022
+export VC_YEAR=2019
 
 if [[ "$DESIRED_CUDA" == 'xpu' ]]; then
     export VC_YEAR=2022

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -12,9 +12,7 @@ body:
     description: |
       Please provide a clear and concise description of what the bug is.
 
-      If relevant, add a minimal example so that we can reproduce the error by running the code. It is very important for the snippet to be as succinct (minimal) as possible, so please take time to trim down any irrelevant code to help us debug efficiently.
-      Your example should be fully self-contained and not rely on any artifact that should be downloaded.
-      For example:
+      If relevant, add a minimal example so that we can reproduce the error by running the code. It is very important for the snippet to be as succinct (minimal) as possible, so please take time to trim down any irrelevant code to help us debug efficiently. We are going to copy-paste your code and we expect to get the same result as you did: avoid any external data, and include the relevant imports, etc. For example:
 
       ```python
       # All necessary imports at the beginning
@@ -28,7 +26,6 @@ body:
       If the code is too long (hopefully, it isn't), feel free to put it in a public gist and link it in the issue: https://gist.github.com.
 
       Please also paste or describe the results you observe instead of the expected results. If you observe an error, please paste the error message including the **full** traceback of the exception. It may be relevant to wrap error messages in ```` ```triple quotes blocks``` ````.
-      If your issue is related to numerical accuracy or reproducibility, please read the [numerical accuracy](https://docs.pytorch.org/docs/stable/notes/numerical_accuracy.html) and [reproducibility](https://docs.pytorch.org/docs/stable/notes/randomness.html) notes. If the difference is not expected as described in these documents, please provide appropriate justification on why one result is wrong and the other is correct.
     placeholder: |
       A clear and concise description of what the bug is.
 

.github/ISSUE_TEMPLATE/disable-ci-jobs.md

@@ -5,7 +5,7 @@ title: "DISABLED [WORKFLOW_NAME] / [PLATFORM_NAME] / [JOB_NAME]"
 labels: "module: ci"
 ---
 
-> For example, DISABLED pull / win-vs2022-cpu-py3 / test (default). Once
+> For example, DISABLED pull / win-vs2019-cpu-py3 / test (default). Once
 > created, the job will be disabled within 15 minutes. You can check the
 > list of disabled jobs at https://ossci-metrics.s3.amazonaws.com/disabled-jobs.json
 

.github/ISSUE_TEMPLATE/release-feature-request.yml

@@ -1,111 +0,0 @@
-name: 🚀 Release highlight for proposed Feature
-description: Submit a Release highlight for proposed Feature
-labels: ["release-feature-request"]
-
-body:
-- type: textarea
-  attributes:
-    label: Release highlight for proposed Feature
-    description: >
-      Example: “A torch.special module, analogous to SciPy's special module.”
-- type: input
-  id: contact
-  attributes:
-    label: Point(s) of contact
-    description: How can we get in touch with you if we need more info?
-    placeholder: ex. github username
-  validations:
-    required: false
-- type: dropdown
-  attributes:
-    label: Release Mode (pytorch/pytorch features only)
-    description: |
-      If "out-of-tree", please include the GH repo name
-    options:
-      - In-tree
-      - Out-of-tree
-  validations:
-    required: true
-- type: textarea
-  attributes:
-    label: Out-Of-Tree Repo
-    description: >
-      please include the GH repo name
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: Description and value to the user
-    description: >
-      Please provide a brief description of the feature and how it will benefit the user.
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: Link to design doc, GitHub issues, past submissions, etc
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: What feedback adopters have provided
-    description: >
-      Please list users/teams that have tried the feature and provided feedback. If that feedback motivated material changes (API, doc, etc..), a quick overview of the changes and the status (planned, in progress, implemented) would be helpful as well.
-  validations:
-    required: false
-- type: dropdown
-  attributes:
-    label: Plan for documentations / tutorials
-    description: |
-      Select One of the following options
-    options:
-      - Tutorial exists
-      - Will submit a PR to pytorch/tutorials
-      - Will submit a PR to a repo
-      - Tutorial is not needed
-  validations:
-    required: true
-- type: textarea
-  attributes:
-    label: Additional context for tutorials
-    description: >
-      Please provide a link for existing tutorial or link to a repo or context for why tutorial is not needed.
-  validations:
-    required: false
-- type: dropdown
-  attributes:
-    label: Marketing/Blog Coverage
-    description: |
-      Are you requesting feature Inclusion in the release blogs?
-    options:
-      - "Yes"
-      - "No"
-  validations:
-    required: true
-- type: textarea
-  attributes:
-    label: Are you requesting other marketing assistance with this feature?
-    description: >
-      E.g. supplementary blogs, social media amplification, etc.
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: Release Version
-    description: >
-      Please include release version for marketing coverage.
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: OS / Platform / Compute Coverage
-    description: >
-      Please list the platforms supported by the proposed feature. If the feature supports all the platforms, write "all". Goal of this section is to clearly share if this feature works in all PyTorch configurations or is it limited to only certain platforms/configurations (e.g. CPU only, GPU only, Linux only, etc...)
-  validations:
-    required: false
-- type: textarea
-  attributes:
-    label: Testing Support (CI, test cases, etc..)
-    description: >
-      Please provide an overview of test coverage. This includes unit testing and integration testing, but if E2E validation testing has been done to show that the feature works for a certain set of use cases or models please mention that as well.
-  validations:
-    required: false

.github/actionlint.yaml

@@ -45,7 +45,6 @@ self-hosted-runner:
     - windows.g5.4xlarge.nvidia.gpu
     # Windows ARM64 runners
     - windows-11-arm64
-    - windows-11-arm64-preview
     # Organization-wide AMD-hosted runners
     # MI2xx runners
     - linux.rocm.gpu

.github/actions/reuse-old-whl/action.yml

@@ -1,38 +0,0 @@
-name: Reuse old wheel if possible
-
-description:
-  Reuse old wheel if possible
-
-inputs:
-  build-environment:
-    description: Build environment
-    required: true
-  run-id:
-    description: Workflow run ID
-    required: true
-  github-token:
-    description: GitHub token
-    required: true
-
-outputs:
-  reuse:
-    description: Whether the wheel is reused or not
-    value: ${{ steps.check-file-changes.outputs.reuse }}
-
-runs:
-  using: composite
-
-  steps:
-    # Check out pytorch with fetch depth 0
-    - name: Check file changes
-      id: check-file-changes
-      shell: bash
-      continue-on-error: true
-      env:
-        GITHUB_TOKEN: ${{ inputs.github-token }}
-      run: |
-        set -x
-        python3 ${GITHUB_ACTION_PATH}/reuse_old_whl.py \
-          --build-environment "${{ inputs.build-environment }}" \
-          --run-id "${{ inputs.run-id }}" \
-          --github-ref "${{ github.ref }}"

.github/actions/reuse-old-whl/reuse_old_whl.py

@@ -1,351 +0,0 @@
-import argparse
-import os
-import subprocess
-from functools import lru_cache
-from pathlib import Path
-from typing import Any, cast, Optional, Union
-
-import requests
-
-
-FORCE_REBUILD_LABEL = "ci-force-rebuild"
-
-
-@lru_cache
-def get_merge_base() -> str:
-    merge_base = subprocess.check_output(
-        ["git", "merge-base", "HEAD", "origin/main"],
-        text=True,
-        stderr=subprocess.DEVNULL,
-    ).strip()
-    # Remove this when we turn this off for the main branch
-    if merge_base == get_head_sha():
-        print("Merge base is the same as HEAD, using HEAD^")
-        merge_base = subprocess.check_output(
-            ["git", "rev-parse", "HEAD^"],
-            text=True,
-            stderr=subprocess.DEVNULL,
-        ).strip()
-    print(f"Merge base: {merge_base}")
-    return merge_base
-
-
-@lru_cache
-def get_head_sha() -> str:
-    sha = subprocess.check_output(
-        ["git", "rev-parse", "HEAD"],
-        text=True,
-        stderr=subprocess.DEVNULL,
-    ).strip()
-    return sha
-
-
-def is_main_branch() -> bool:
-    return False
-    # Testing on main branch for now
-    # print(
-    #     f"Checking if we are on main branch: merge base {get_merge_base()}, head {get_head_sha()}"
-    # )
-    # return get_merge_base() == get_head_sha()
-
-
-def query_github_api(url: str) -> Any:
-    headers = {
-        "Accept": "application/vnd.github.v3+json",
-        "Authorization": f"Bearer {os.environ['GITHUB_TOKEN']}",
-    }
-    response = requests.get(url, headers=headers)
-    return response.json()
-
-
-@lru_cache
-def check_labels_for_pr() -> bool:
-    # Check if the current commit is part of a PR and if it has the
-    # FORCE_REBUILD_LABEL
-    head_sha = get_head_sha()
-    url = f"https://api.github.com/repos/pytorch/pytorch/commits/{head_sha}/pulls"
-    response = query_github_api(url)
-
-    print(
-        f"Found {len(response)} PRs for commit {head_sha}: {[pr['number'] for pr in response]}"
-    )
-    for pr in response:
-        labels = pr.get("labels", [])
-        for label in labels:
-            if label["name"] == FORCE_REBUILD_LABEL:
-                print(f"Found label {FORCE_REBUILD_LABEL} in PR {pr['number']}.")
-                return True
-    return False
-
-
-def check_issue_open() -> bool:
-    # Check if issue #153759 is open.  This is the config issue for quickly
-    # forcing everyone to build
-    url = "https://api.github.com/repos/pytorch/pytorch/issues/153759"
-    response = query_github_api(url)
-    if response.get("state") == "open":
-        print("Issue #153759 is open.")
-        return True
-    else:
-        print("Issue #153759 is not open.")
-        return False
-
-
-def get_workflow_id(run_id: str) -> Optional[str]:
-    # Get the workflow ID that corresponds to the file for the run ID
-    url = f"https://api.github.com/repos/pytorch/pytorch/actions/runs/{run_id}"
-    response = query_github_api(url)
-    if "workflow_id" in response:
-        print(f"Found workflow ID for run ID {run_id}: {response['workflow_id']}")
-        return cast(str, response["workflow_id"])
-    else:
-        print("No workflow ID found.")
-        return None
-
-
-def ok_changed_file(file: str) -> bool:
-    # Return true if the file is in the list of allowed files to be changed to
-    # reuse the old whl
-    if (
-        file.startswith("torch/")
-        and file.endswith(".py")
-        and not file.startswith("torch/csrc/")
-    ):
-        return True
-    if file.startswith("test/") and file.endswith(".py"):
-        return True
-    return False
-
-
-def check_changed_files(sha: str) -> bool:
-    # Return true if all the changed files are in the list of allowed files to
-    # be changed to reuse the old whl
-
-    # Removing any files is not allowed since rysnc will not remove files
-    removed_files = (
-        subprocess.check_output(
-            ["git", "diff", "--name-only", sha, "HEAD", "--diff-filter=D"],
-            text=True,
-            stderr=subprocess.DEVNULL,
-        )
-        .strip()
-        .split()
-    )
-    if removed_files:
-        print(
-            f"Removed files between {sha} and HEAD: {removed_files}, cannot reuse old whl"
-        )
-        return False
-
-    changed_files = (
-        subprocess.check_output(
-            ["git", "diff", "--name-only", sha, "HEAD"],
-            text=True,
-            stderr=subprocess.DEVNULL,
-        )
-        .strip()
-        .split()
-    )
-    print(f"Checking changed files between {sha} and HEAD:")
-    for file in changed_files:
-        if not ok_changed_file(file):
-            print(f"  File {file} is not allowed to be changed.")
-            return False
-        else:
-            print(f"  File {file} is allowed to be changed.")
-    return True
-
-
-def find_old_whl(workflow_id: str, build_environment: str, sha: str) -> bool:
-    # Find the old whl on s3 and download it to artifacts.zip
-    if build_environment is None:
-        print("BUILD_ENVIRONMENT is not set.")
-        return False
-    print(f"SHA: {sha}, workflow_id: {workflow_id}")
-
-    workflow_runs = query_github_api(
-        f"https://api.github.com/repos/pytorch/pytorch/actions/workflows/{workflow_id}/runs?head_sha={sha}&branch=main&per_page=100"
-    )
-    if workflow_runs.get("total_count", 0) == 0:
-        print("No workflow runs found.")
-        return False
-    for run in workflow_runs.get("workflow_runs", []):
-        # Look in s3 for the old whl
-        run_id = run["id"]
-        try:
-            url = f"https://gha-artifacts.s3.amazonaws.com/pytorch/pytorch/{run_id}/{build_environment}/artifacts.zip"
-            print(f"Checking for old whl at {url}")
-            response = requests.get(
-                url,
-            )
-            if response.status_code == 200:
-                with open("artifacts.zip", "wb") as f:
-                    f.write(response.content)
-                    print(f"Found old whl file from s3: {url}")
-                    return True
-        except requests.RequestException as e:
-            print(f"Error checking for old whl: {e}")
-            continue
-    return False
-
-
-def unzip_artifact_and_replace_files() -> None:
-    # Unzip the artifact and replace files
-    subprocess.check_output(
-        ["unzip", "-o", "artifacts.zip", "-d", "artifacts"],
-    )
-    os.remove("artifacts.zip")
-
-    head_sha = get_head_sha()
-
-    # Rename wheel into zip
-    wheel_path = Path("artifacts/dist").glob("*.whl")
-    for path in wheel_path:
-        # Should be of the form torch-2.0.0+git1234567-cp37-etc.whl
-        # Should usually be the merge base sha but for the ones that didn't do
-        # the replacement, it won't be.  Can probably change it to just be merge
-        # base later
-        old_version = f"+git{path.stem.split('+')[1].split('-')[0][3:]}"
-        new_version = f"+git{head_sha[:7]}"
-
-        def rename_to_new_version(file: Union[str, Path]) -> None:
-            # Rename file with old_version to new_version
-            subprocess.check_output(
-                ["mv", file, str(file).replace(old_version, new_version)]
-            )
-
-        def change_content_to_new_version(file: Union[str, Path]) -> None:
-            # Check if is a file
-            if os.path.isdir(file):
-                return
-            # Replace the old version in the file with the new version
-            with open(file) as f:
-                content = f.read()
-                content = content.replace(old_version, new_version)
-            with open(file, "w") as f:
-                f.write(content)
-
-        zip_path = path.with_suffix(".zip")
-        os.rename(path, zip_path)
-        old_stem = zip_path.stem
-        # Unzip the wheel
-        subprocess.check_output(
-            ["unzip", "-o", zip_path, "-d", f"artifacts/dist/{old_stem}"],
-        )
-
-        # Remove the old wheel (which is now a zip file)
-        os.remove(zip_path)
-
-        # Copy python files into the artifact
-        subprocess.check_output(
-            ["rsync", "-avz", "torch", f"artifacts/dist/{old_stem}"],
-        )
-
-        change_content_to_new_version(f"artifacts/dist/{old_stem}/torch/version.py")
-
-        for file in Path(f"artifacts/dist/{old_stem}").glob(
-            "*.dist-info/**",
-        ):
-            change_content_to_new_version(file)
-
-        rename_to_new_version(f"artifacts/dist/{old_stem}")
-        new_stem = old_stem.replace(old_version, new_version)
-
-        for file in Path(f"artifacts/dist/{new_stem}").glob(
-            "*.dist-info",
-        ):
-            rename_to_new_version(file)
-
-        # Zip the wheel back
-        subprocess.check_output(
-            ["zip", "-r", f"{new_stem}.zip", "."],
-            cwd=f"artifacts/dist/{new_stem}",
-        )
-
-        subprocess.check_output(
-            [
-                "mv",
-                f"artifacts/dist/{new_stem}/{new_stem}.zip",
-                f"artifacts/dist/{new_stem}.whl",
-            ],
-        )
-
-        # Remove the extracted folder
-        subprocess.check_output(
-            ["rm", "-rf", f"artifacts/dist/{new_stem}"],
-        )
-
-    # Rezip the artifact
-    subprocess.check_output(["zip", "-r", "artifacts.zip", "."], cwd="artifacts")
-    subprocess.check_output(
-        ["mv", "artifacts/artifacts.zip", "."],
-    )
-    return None
-
-
-def set_output() -> None:
-    # Disable for now so we can monitor first
-    # pass
-    if os.getenv("GITHUB_OUTPUT"):
-        with open(str(os.getenv("GITHUB_OUTPUT")), "a") as env:
-            print("reuse=true", file=env)
-    else:
-        print("::set-output name=reuse::true")
-
-
-def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(description="Check for old whl files.")
-    parser.add_argument("--run-id", type=str, required=True, help="Workflow ID")
-    parser.add_argument(
-        "--build-environment", type=str, required=True, help="Build environment"
-    )
-    parser.add_argument(
-        "--github-ref",
-        type=str,
-    )
-    return parser.parse_args()
-
-
-def can_reuse_whl(args: argparse.Namespace) -> bool:
-    # if is_main_branch() or (
-    #     args.github_ref
-    #     and any(
-    #         args.github_ref.startswith(x)
-    #         for x in ["refs/heads/release", "refs/tags/v", "refs/heads/main"]
-    #     )
-    # ):
-    #     print("On main branch or release branch, rebuild whl")
-    #     return False
-
-    if check_labels_for_pr():
-        print(f"Found {FORCE_REBUILD_LABEL} label on PR, rebuild whl")
-        return False
-
-    if check_issue_open():
-        print("Issue #153759 is open, rebuild whl")
-        return False
-
-    if not check_changed_files(get_merge_base()):
-        print("Cannot use old whl due to the changed files, rebuild whl")
-        return False
-
-    workflow_id = get_workflow_id(args.run_id)
-    if workflow_id is None:
-        print("No workflow ID found, rebuild whl")
-        return False
-
-    if not find_old_whl(workflow_id, args.build_environment, get_merge_base()):
-        print("No old whl found, rebuild whl")
-        # TODO: go backwards from merge base to find more runs
-        return False
-
-    return True
-
-
-if __name__ == "__main__":
-    args = parse_args()
-
-    if can_reuse_whl(args):
-        print("Reusing old whl")
-        unzip_artifact_and_replace_files()
-        set_output()

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

@@ -29,13 +29,13 @@ runs:
       if: always()
       shell: bash
       run: |
-        timeout 30 xpu-smi discovery || true
+        xpu-smi discovery
 
     - name: Runner health check GPU count
       if: always()
       shell: bash
       run: |
-        ngpu=$(timeout 30 xpu-smi discovery | grep -c -E 'Device Name')
+        ngpu=$(xpu-smi discovery | grep -c -E 'Device Name')
         msg="Please file an issue on pytorch/pytorch reporting the faulty runner. Include a link to the runner logs so the runner can be identified"
         if [[ $ngpu -eq 0 ]]; then
           echo "Error: Failed to detect any GPUs on the runner"

.github/actions/upload-utilization-stats/action.yml

@@ -1,6 +1,6 @@
 name: upload-utilization-stats
 
-description: Upload utilization stats to artifacts.
+description: Upload utilization stats to artifacts
 
 inputs:
     workflow_run_id:
@@ -23,17 +23,6 @@ inputs:
       type: string
       description: 'the job name of the test'
       required: True
-    local_path:
-      type: string
-      description: 'the local path to the utilization stats file'
-      required: False
-      default: ''
-    artifact_prefix:
-      type: string
-      description: |
-          'the prefix of the raw utilization data, for data stored in zip file, this is the prefix of the parent zip file'
-      default: ""
-      required: False
 
 runs:
   using: composite
@@ -46,8 +35,6 @@ runs:
         echo "workflow_Name: ${{inputs.workflow_name}}"
         echo "job_id: ${{inputs.job_id}}"
         echo "job_name:  ${{inputs.job_name}}"
-        echo "artifact_prefix: ${{inputs.artifact_prefix}}"
-        python3 --version
     - uses: nick-fields/retry@v3.0.0
       name: Setup dependencies
       with:
@@ -66,6 +53,4 @@ runs:
           --workflow-name "${{inputs.workflow_name}}" \
           --workflow-run-attempt "${{inputs.workflow_attempt}}" \
           --job-id "${{inputs.job_id}}" \
-          --job-name "${{inputs.job_name}}" \
-          --local-path "${{inputs.local_path}}" \
-          --artifact-prefix "${{inputs.artifact_prefix}}"
+          --job-name "${{inputs.job_name}}"

.github/ci_commit_pins/audio.txt

@@ -1 +1 @@
-4cb7f57d31b0b288696f09b89e890e5fac092eed
+ea5de17755d657508c84c4dce8970b614008adcf

.github/ci_commit_pins/torchbench.txt

@@ -1 +1 @@
-e03a63be43e33596f7f0a43b0f530353785e4a59
+373ffb19dc470f4423a3176a4133f8f4b3cdb5bd

.github/ci_commit_pins/vision.txt

@@ -1 +1 @@
-966da7e46f65d6d49df3e31214470a4fe5cc8e66
+d23a6e1664d20707c11781299611436e1f0c104f

.github/ci_commit_pins/xla.txt

@@ -1 +1 @@
-edc1a882d872dd7f1362e4312fd045a1d81b3355
+8d9e34b352af09c81ff8df448fd27f9c4aae1382

.github/merge_rules.yaml

@@ -123,8 +123,6 @@
   - torch/*docs.py
   approved_by:
   - svekars
-  - sekyondaMeta
-  - AlannaBurke
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -400,10 +398,7 @@
   - torch/_inductor/codegen/cpp_micro_gemm.py
   - torch/_inductor/codegen/cpp_template_kernel.py
   - torch/_inductor/codegen/cpp_template.py
-  - torch/_inductor/codegen/cpp_bmm_template.py
   - torch/_inductor/codegen/cpp_gemm_template.py
-  - torch/_inductor/codegen/cpp_grouped_gemm_template.py
-  - torch/_inductor/codegen/cpp_flex_attention_template.py
   - torch/csrc/inductor/cpp_prefix.h
   - test/inductor/test_mkldnn_pattern_matcher.py
   - test/inductor/test_cpu_repro.py
@@ -411,7 +406,6 @@
   - test/inductor/test_cpu_select_algorithm.py
   - aten/src/ATen/cpu/**
   - aten/src/ATen/native/quantized/cpu/**
-  - aten/src/ATen/test/vec_test_all_types.*
   - test/quantization/core/test_quantized_op.py
   - torch/ao/quantization/quantizer/x86_inductor_quantizer.py
   - test/quantization/pt2e/test_x86inductor_quantizer.py
@@ -419,7 +413,6 @@
   - leslie-fang-intel
   - jgong5
   - EikanWang
-  - CaoE
   mandatory_checks_name:
   - EasyCLA
   - Lint

.github/pytorch-probot.yml

@@ -28,7 +28,6 @@ ciflow_push_tags:
 - ciflow/op-benchmark
 - ciflow/pull
 - ciflow/h100
-- ciflow/h100-distributed
 retryable_workflows:
 - pull
 - trunk

.github/requirements/README.md

@@ -11,6 +11,16 @@ jobs, but it also allows them to be cached properly to improve CI
 reliability.
 
 The list of support files are as follows:
+
+* Conda:
+  * conda-env-iOS. This is used by iOS build and test jobs to setup the
+    conda environment
+  * conda-env-macOS-ARM64. This is used by MacOS (m1, arm64) build and
+    test jobs to setup the conda environment
+  * conda-env-Linux-X64. This is used by Linux buck build and test jobs
+    to setup the conda environment
 * Pip:
+  * pip-requirements-iOS.txt. This is used by iOS build and test jobs to
+    setup the pip environment
   * pip-requirements-macOS.txt. This is used by MacOS build and test jobs to
     setup the pip environment

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

@@ -0,0 +1,8 @@
+cmake=3.22.*
+mkl=2022.1.0
+mkl-include=2022.1.0
+ninja=1.10.2
+numpy=1.23.3
+pyyaml=6.0
+setuptools=72.1.0
+typing-extensions=4.11.0

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

@@ -0,0 +1,7 @@
+blas=1.0
+cmake=3.22.1
+ninja=1.10.2
+numpy=1.23.3
+pyyaml=6.0
+setuptools=72.1.0
+typing-extensions=4.11.0

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

@@ -1,6 +1,22 @@
+numpy=1.22.3
+pyyaml=6.0
+setuptools=72.1.0
+cmake=3.22.*
+typing-extensions=4.11.0
+dataclasses=0.8
+pip=22.2.2
+pillow=10.0.1
+pkg-config=0.29.2
+wheel=0.37.1
+# NB: This is intentionally held back because anaconda main doesn't
+# have updated expecttest, but you don't /need/ the updated version
+# to run the tests.  In the meantime I need to figure out how to
+# cajole anaconda into updating, or get the package from pypi instead...
+expecttest=0.1.3
+
 # Not pinning certifi so that we can always get the latest certificates
 certifi
-pip=23.2.1
-pkg-config=0.29.2
-setuptools=72.1.0
-wheel=0.37.1
+
+# Cross-compiling arm64 from x86-64 picks up 1.40.0 while testing on arm64
+# itself only has up to 1.39.0 from upstream conda. Both work though
+libuv>=1.39.0,<=1.40.0

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

@@ -0,0 +1,4 @@
+# iOS simulator requirements
+coremltools==5.0b5
+protobuf==3.20.2
+optree==0.13.0

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

@@ -1,34 +1,33 @@
 boto3==1.35.42
-cmake==3.25.*
+hypothesis==6.56.4
 expecttest==0.3.0
 fbscribelogger==0.1.7
-filelock==3.6.0
-hypothesis==6.56.4
 librosa>=0.6.2
 mpmath==1.3.0
 networkx==2.8.7
-ninja==1.10.2.4
-numba==0.59.0
-numpy==1.26.4
+# Use numba-0.49.1 or older on Intel Macs, but 0.56.0 on M1 machines, as older numba is not available
+numba==0.56.0; platform_machine == "arm64"
+numba<=0.49.1; platform_machine != "arm64"
 opt-einsum>=3.3
-optree==0.13.0
-packaging==23.1
-parameterized==0.8.1
-pillow==10.3.0
-protobuf==5.29.4
 psutil==5.9.1
+nvidia-ml-py==11.525.84
+packaging==23.1
 pygments==2.15.0
-pytest-cpp==2.3.0
-pytest-flakefinder==1.1.0
-pytest-rerunfailures==10.3
-pytest-subtests==0.13.1
-pytest-xdist==3.3.1
 pytest==7.3.2
-pyyaml==6.0.2
-scipy==1.12.0
+pytest-xdist==3.3.1
+pytest-rerunfailures==10.3
+pytest-flakefinder==1.1.0
+pytest-subtests==0.13.1
+scipy==1.10.1
 sympy==1.13.3
-tensorboard==2.13.0
-typing-extensions==4.12.2
 unittest-xml-reporting<=3.2.0,>=2.0.0
 xdoctest==1.1.0
+filelock==3.6.0
+pytest-cpp==2.3.0
 z3-solver==4.12.2.0
+tensorboard==2.13.0
+optree==0.13.0
+# 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
+parameterized==0.8.1

.github/scripts/build_triton_wheel.py

@@ -65,7 +65,6 @@ def build_triton(
     with TemporaryDirectory() as tmpdir:
         triton_basedir = Path(tmpdir) / "triton"
         triton_pythondir = triton_basedir / "python"
-
         triton_repo = "https://github.com/openai/triton"
         if device == "rocm":
             triton_pkg_name = "pytorch-triton-rocm"
@@ -102,19 +101,11 @@ def build_triton(
             )
             print("ROCm libraries setup for triton installation...")
 
-        # old triton versions have setup.py in the python/ dir,
-        # new versions have it in the root dir.
-        triton_setupdir = (
-            triton_basedir
-            if (triton_basedir / "setup.py").exists()
-            else triton_pythondir
-        )
-
         check_call(
-            [sys.executable, "setup.py", "bdist_wheel"], cwd=triton_setupdir, env=env
+            [sys.executable, "setup.py", "bdist_wheel"], cwd=triton_pythondir, env=env
         )
 
-        whl_path = next(iter((triton_setupdir / "dist").glob("*.whl")))
+        whl_path = next(iter((triton_pythondir / "dist").glob("*.whl")))
         shutil.copy(whl_path, Path.cwd())
 
         if device == "rocm":

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

@@ -28,12 +28,12 @@ def main() -> None:
     issue = repo.get_issue(issue_number)
     issue_labels = issue.labels
     docathon_label_present = any(
-        label.name == "docathon-h1-2025" for label in issue_labels
+        label.name == "docathon-h1-2024" for label in issue_labels
     )
 
     # if the issue has a docathon label, add all labels from the issue to the PR.
     if not docathon_label_present:
-        print("The 'docathon-h1-2025' label is not present in the issue.")
+        print("The 'docathon-h1-2024' label is not present in the issue.")
         return
     pull_request_labels = pull_request.get_labels()
     pull_request_label_names = [label.name for label in pull_request_labels]

.github/scripts/filter_test_configs.py

@@ -80,7 +80,7 @@ def parse_args() -> Any:
     parser.add_argument(
         "--job-name",
         type=str,
-        help="the name of the current job, i.e. linux-jammy-py3.8-gcc7 / build",
+        help="the name of the current job, i.e. linux-focal-py3.8-gcc7 / build",
     )
     parser.add_argument("--pr-number", type=str, help="the pull request number")
     parser.add_argument("--tag", type=str, help="the associated tag if it exists")

.github/scripts/generate_binary_build_matrix.py

@@ -21,7 +21,7 @@ CUDA_STABLE = "12.6"
 CUDA_ARCHES_FULL_VERSION = {
     "11.8": "11.8.0",
     "12.6": "12.6.3",
-    "12.8": "12.8.1",
+    "12.8": "12.8.0",
 }
 CUDA_ARCHES_CUDNN_VERSION = {
     "11.8": "9",
@@ -65,29 +65,27 @@ PYTORCH_EXTRA_INSTALL_REQUIREMENTS = {
         "nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
     "12.8": (
-        "nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | "
         "nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | "
-        "nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'"
+        "nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | "
+        "nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'"
     ),
     "xpu": (
         "intel-cmplr-lib-rt==2025.1.1 | "

.github/scripts/lintrunner.sh

@@ -31,9 +31,6 @@ python3 -m tools.pyi.gen_pyi \
     --deprecated-functions-path "tools/autograd/deprecated.yaml"
 python3 torch/utils/data/datapipes/gen_pyi.py
 
-# Also check generated pyi files
-find torch -name '*.pyi' -exec git add --force -- "{}" +
-
 RC=0
 # Run lintrunner on all files
 if ! lintrunner --force-color --tee-json=lint.json ${ADDITIONAL_LINTRUNNER_ARGS} 2> /dev/null; then
@@ -44,9 +41,6 @@ if ! lintrunner --force-color --tee-json=lint.json ${ADDITIONAL_LINTRUNNER_ARGS}
     RC=1
 fi
 
-# Unstage temporally added pyi files
-find torch -name '*.pyi' -exec git restore --staged -- "{}" +
-
 # Use jq to massage the JSON lint output into GitHub Actions workflow commands.
 jq --raw-output \
     '"::\(if .severity == "advice" or .severity == "disabled" then "warning" else .severity end) file=\(.path),line=\(.line),col=\(.char),title=\(.code) \(.name)::" + (.description | gsub("\\n"; "%0A"))' \

.github/scripts/s390x-ci/self-hosted-builder/fs/usr/bin/actions-runner

@@ -27,9 +27,6 @@ unset ACCESS_TOKEN
 # it does one job, stops and unregisters
 registration_token=$(jq --raw-output .token "$token_file")
 
-# workaround for https://gitlab.com/qemu-project/qemu/-/issues/2600
-export DOTNET_EnableWriteXorExecute=0
-
 ./config.sh \
         --unattended \
         --ephemeral \
@@ -47,5 +44,8 @@ rm -f "$token_file"
 # and it doesn't work for non-root user
 source venv/bin/activate
 
+# workaround for https://gitlab.com/qemu-project/qemu/-/issues/2600
+export DOTNET_EnableWriteXorExecute=0
+
 # Run one job.
 ./run.sh

.github/scripts/trymerge.py

@@ -939,12 +939,6 @@ class GitHubPR:
                     summary=None,
                 )
 
-        # Making an exception for Apply lint auggestions/autoformat because the
-        # bot adds a merged label -> triggers workflow -> sometimes needs
-        # approval -> is read as failure, which results in a blocked merge, but
-        # this workflow doesn't provide mergability info
-        self.conclusions.pop("Apply lint suggestions", None)
-
         return self.conclusions
 
     def get_authors(self) -> dict[str, str]:

.github/templates/windows_binary_build_workflow.yml.j2

@@ -76,7 +76,7 @@ jobs:
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: get-label-type
     {%- if os == "windows-arm64" %}
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     {%- else %}
     {%- if branches == "nightly" %}
     runs-on: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge"
@@ -102,12 +102,23 @@ jobs:
         run: |
           mkdir "%NIGHTLIES_PYTORCH_ROOT%"
           mkdir "%PYTORCH_FINAL_PACKAGE_DIR%"
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Bootstrap Build Tools
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
-      - name: Git checkout PyTorch
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
+      - name: Bootstrap Git
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
+      - name: Git checkout PyTorch - recursive
         uses: actions/checkout@v4
         with:
           path: "pytorch"
@@ -161,7 +172,7 @@ jobs:
       - !{{ config["build_name"] }}-build
       - get-label-type
 {%- if os == "windows-arm64" %}
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
 {%- else %}
 {%- if config["gpu_arch_type"] == "cuda" %}
 {%- if branches == "nightly" %}
@@ -187,11 +198,18 @@ jobs:
           echo BINARY_ENV_FILE=%RUNNER_TEMP%/env>> %GITHUB_ENV%
           echo PYTORCH_FINAL_PACKAGE_DIR=%RUNNER_TEMP%/artifacts>> %GITHUB_ENV%
           echo WIN_PACKAGE_WORK_DIR=%RUNNER_TEMP%>> %GITHUB_ENV%
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Populate binary env
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
       - name: Git checkout PyTorch
         uses: actions/checkout@v4
         with:
@@ -205,6 +223,10 @@ jobs:
         shell: cmd
         run: |
           "pytorch/.ci/pytorch/windows/arm64/bootstrap_python.bat"
+      - name: Bootstrap Build Tools
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
       - name: Bootstrap Rust
         shell: cmd
         run: |

.github/workflows/_link_check.yml

@@ -7,25 +7,29 @@ on:
       ref:
         type: string
         required: true
+      run-url-lint:
+        type: boolean
+        required: false
+        default: false
 
 jobs:
   lint-urls:
-    if: ${{ github.event_name != 'pull_request' || !contains(github.event.pull_request.labels.*.name, 'skip-url-lint') }}
+    if: ${{ inputs.run-url-lint }}
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
     with:
       timeout: 120
       runner: ${{ inputs.runner }}linux.2xlarge
-      docker-image: ci-image:pytorch-linux-jammy-linter
+      docker-image: pytorch-linux-focal-linter
       fetch-depth: 0
       submodules: false
       ref: ${{ inputs.ref }}
       script: |
         ./scripts/lint_urls.sh $(
-          if [ "${{ github.event_name }}" = "pull_request" ]; then
-            echo "${{ github.event.pull_request.base.sha }}" "${{ github.event.pull_request.head.sha }}"
-          else
-            echo "${{ github.event.before }}" "${{ github.sha }}"
-          fi
+          { [ "${{ github.event_name }}" = "pull_request" ] \
+              && git diff --name-only "${{ github.event.pull_request.base.sha }}...${{ github.event.pull_request.head.sha }}"; } \
+          || \
+          { [ "${{ github.event_name }}" = "push" ] \
+              && git diff --name-only "${{ github.event.before }}...${{ github.sha }}"; }
         ) || {
           echo
           echo "URL lint failed."
@@ -40,17 +44,17 @@ jobs:
     with:
       timeout: 60
       runner: ${{ inputs.runner }}linux.2xlarge
-      docker-image: ci-image:pytorch-linux-jammy-linter
+      docker-image: pytorch-linux-focal-linter
       fetch-depth: 0
       submodules: false
       ref: ${{ inputs.ref }}
       script: |
         ./scripts/lint_xrefs.sh $(
-          if [ "${{ github.event_name }}" = "pull_request" ]; then
-            echo "${{ github.event.pull_request.base.sha }}" "${{ github.event.pull_request.head.sha }}"
-          else
-            echo "${{ github.event.before }}" "${{ github.sha }}"
-          fi
+          { [ "${{ github.event_name }}" = "pull_request" ] \
+              && git diff --name-only "${{ github.event.pull_request.base.sha }}...${{ github.event.pull_request.head.sha }}"; } \
+          || \
+          { [ "${{ github.event_name }}" = "push" ] \
+              && git diff --name-only "${{ github.event.before }}...${{ github.sha }}"; }
         ) || {
           echo
           echo "Xref lint failed."

.github/workflows/_linux-build.yml

@@ -74,32 +74,6 @@ on:
           Overwrite the number of jobs to use for the build
         required: false
         type: string
-      disable-monitor:
-        description: |
-          Disable utilization monitoring for build job
-        required: false
-        type: boolean
-        default: false
-      monitor-log-interval:
-        description: |
-          Set the interval for the monitor script to log utilization.
-        required: false
-        type: number
-        default: 5
-      monitor-data-collect-interval:
-        description: |
-          Set the interval for the monitor script to collect data.
-        required: false
-        type: number
-        default: 1
-
-      allow-reuse-old-whl:
-        description: |
-          If set, the build try to pull an old wheel from s3 that was built on a
-          commit with no cpp changes from this commit
-        required: false
-        type: boolean
-        default: false
 
     secrets:
       HUGGING_FACE_HUB_TOKEN:
@@ -158,15 +132,6 @@ jobs:
           role-session-name: gha-linux-build
           aws-region: us-east-1
 
-      - name: Check if can use old whl build
-        id: use-old-whl
-        uses: ./.github/actions/reuse-old-whl
-        if: ${{ inputs.allow-reuse-old-whl }}
-        with:
-          build-environment: ${{ inputs.build-environment }}
-          run-id: ${{ github.run_id }}
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-
       - name: Calculate docker image
         id: calculate-docker-image
         uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
@@ -176,7 +141,7 @@ jobs:
 
       - name: Use following to pull public copy of the image
         id: print-ghcr-mirror
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel' && steps.use-old-whl.outputs.reuse != 'true'
+        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         env:
           ECR_DOCKER_IMAGE: ${{ steps.calculate-docker-image.outputs.docker-image }}
         shell: bash
@@ -186,7 +151,7 @@ jobs:
 
       - name: Pull docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel' && steps.use-old-whl.outputs.reuse != 'true'
+        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -211,38 +176,17 @@ jobs:
           selected-test-configs: ${{ inputs.selected-test-configs }}
           job-name: ${{ steps.get-job-id.outputs.job-name }}
 
-      - name: Start monitoring script
-        id: monitor-script
-        if: ${{ !inputs.disable-monitor }}
-        shell: bash
-        continue-on-error: true
-        env:
-          JOB_ID: ${{ steps.get-job-id.outputs.job-id }}
-          JOB_NAME: ${{ steps.get-job-id.outputs.job-name }}
-          WORKFLOW_NAME: ${{ github.workflow }}
-          WORKFLOW_RUN_ID: ${{github.run_id}}
-          MONITOR_LOG_INTERVAL: ${{ inputs.monitor-log-interval }}
-          MONITOR_DATA_COLLECT_INTERVAL: ${{ inputs.monitor-data-collect-interval }}
-        run: |
-          mkdir -p ../../usage_logs
-          python3 -m pip install psutil==5.9.1 dataclasses_json==0.6.7
-          python3 -m tools.stats.monitor \
-          --log-interval "$MONITOR_LOG_INTERVAL" \
-          --data-collect-interval "$MONITOR_DATA_COLLECT_INTERVAL" \
-          > "../../usage_logs/usage_log_build_${JOB_ID}.txt" 2>&1 &
-          echo "monitor-script-pid=${!}" >> "${GITHUB_OUTPUT}"
-
       - name: Download pytest cache
         uses: ./.github/actions/pytest-cache-download
         continue-on-error: true
-        if: inputs.build-environment != 'linux-s390x-binary-manywheel' && steps.use-old-whl.outputs.reuse != 'true'
+        if: inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           cache_dir: .pytest_cache
           job_identifier: ${{ github.workflow }}_${{ inputs.build-environment }}
           s3_bucket: ${{ inputs.s3-bucket }}
 
       - name: Build
-        if: (steps.filter.outputs.is-test-matrix-empty == 'False' || inputs.test-matrix == '') && steps.use-old-whl.outputs.reuse != 'true'
+        if: steps.filter.outputs.is-test-matrix-empty == 'False' || inputs.test-matrix == ''
         id: build
         env:
           BUILD_ENVIRONMENT: ${{ inputs.build-environment }}
@@ -336,23 +280,14 @@ jobs:
           END_TIME=$(date +%s)
           echo "build_time=$((END_TIME - START_TIME))" >> "$GITHUB_OUTPUT"
 
-      - name: Stop monitoring script
-        if: ${{ always() && steps.monitor-script.outputs.monitor-script-pid }}
-        shell: bash
-        continue-on-error: true
-        env:
-          MONITOR_SCRIPT_PID: ${{ steps.monitor-script.outputs.monitor-script-pid }}
-        run: |
-          kill "$MONITOR_SCRIPT_PID"
-
       - name: Archive artifacts into zip
-        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && steps.use-old-whl.outputs.reuse != 'true'
+        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped'
         run: |
           zip -1 -r artifacts.zip dist/ build/custom_test_artifacts build/lib build/bin .additional_ci_files
 
       - name: Store PyTorch Build Artifacts on S3
         uses: seemethere/upload-artifact-s3@baba72d0712b404f646cebe0730933554ebce96a # v5.1.0
-        if: inputs.build-generates-artifacts && (steps.build.outcome != 'skipped' || steps.use-old-whl.outputs.reuse == 'true') && inputs.build-environment != 'linux-s390x-binary-manywheel'
+        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.build-environment != 'linux-s390x-binary-manywheel'
         with:
           name: ${{ inputs.build-environment }}
           retention-days: 14
@@ -362,32 +297,13 @@ jobs:
 
       - name: Store PyTorch Build Artifacts for s390x
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
-        if: inputs.build-generates-artifacts && (steps.build.outcome != 'skipped' || steps.use-old-whl.outputs.reuse == 'true') && inputs.build-environment == 'linux-s390x-binary-manywheel'
+        if: inputs.build-generates-artifacts && steps.build.outcome != 'skipped' && inputs.build-environment == 'linux-s390x-binary-manywheel'
         with:
           name: ${{ inputs.build-environment }}
           retention-days: 14
           if-no-files-found: error
           path: artifacts.zip
 
-      - name: copy logs
-        shell: bash
-        if: ${{ always() && steps.build.outcome != 'skipped' && !inputs.disable-monitor && inputs.build-environment != 'linux-s390x-binary-manywheel'}}
-        continue-on-error: true
-        run: |
-          rm -f ./usage_logs
-          mkdir -p ./usage_logs
-          cp ../../usage_logs/usage_log_build_*.txt ./usage_logs/
-
-      - name: Upload raw usage log to s3
-        if: ${{ always() && steps.build.outcome != 'skipped' && !inputs.disable-monitor && inputs.build-environment != 'linux-s390x-binary-manywheel'}}
-        uses: seemethere/upload-artifact-s3@v5
-        with:
-          s3-prefix: |
-            ${{ github.repository }}/${{ github.run_id }}/${{ github.run_attempt }}/artifact
-          retention-days: 14
-          if-no-files-found: warn
-          path: usage_logs/usage_log_build_*.txt
-
       - name: Upload sccache stats
         if: steps.build.outcome != 'skipped' && inputs.build-environment != 'linux-s390x-binary-manywheel'
         uses: ./.github/actions/upload-sccache-stats
@@ -395,18 +311,6 @@ jobs:
           github-token: ${{ secrets.GITHUB_TOKEN }}
           build-time: ${{ steps.build.outputs.build_time }}
 
-      - name: Upload utilization stats
-        if: ${{ always() && steps.build.outcome != 'skipped' && !inputs.disable-monitor && inputs.build-environment != 'linux-s390x-binary-manywheel' }}
-        continue-on-error: true
-        uses: ./.github/actions/upload-utilization-stats
-        with:
-          job_id: ${{ steps.get-job-id.outputs.job-id }}
-          job_name: ${{ steps.get-job-id.outputs.job-name }}
-          workflow_name: ${{ github.workflow }}
-          workflow_run_id: ${{github.run_id}}
-          workflow_attempt: ${{github.run_attempt}}
-          artifact_prefix: usage_log_build_${{ steps.get-job-id.outputs.job-id }}
-
       - name: Teardown Linux
         uses: pytorch/test-infra/.github/actions/teardown-linux@main
         if: always() && inputs.build-environment != 'linux-s390x-binary-manywheel'

.github/workflows/_linux-test.yml

@@ -376,7 +376,7 @@ jobs:
       - name: Upload pytest cache if tests failed
         uses: ./.github/actions/pytest-cache-upload
         continue-on-error: true
-        if: failure() && steps.test.conclusion && steps.test.conclusion == 'failure' && inputs.build-environment != 'linux-s390x-binary-manywheel'
+        if: failure() && steps.test.conclusion && steps.test.conclusion == 'failure'
         with:
           cache_dir: .pytest_cache
           shard: ${{ matrix.shard }}
@@ -431,7 +431,7 @@ jobs:
           path: ./**/core.[1-9]*
 
       - name: Upload utilization stats
-        if: ${{ always() && steps.test.conclusion && steps.test.conclusion != 'skipped' && !inputs.disable-monitor && inputs.build-environment != 'linux-s390x-binary-manywheel' }}
+        if: ${{ always() && steps.test.conclusion && steps.test.conclusion != 'skipped' && !inputs.disable-monitor }}
         continue-on-error: true
         uses: ./.github/actions/upload-utilization-stats
         with:

.github/workflows/_mac-build.yml

@@ -30,7 +30,7 @@ on:
       python-version:
         required: false
         type: string
-        default: "3.12"
+        default: "3.9"
         description: |
           The python version to be used. Will be 3.9 by default
       test-matrix:
@@ -85,9 +85,8 @@ jobs:
         uses: pytorch/test-infra/.github/actions/setup-miniconda@main
         with:
           python-version: ${{ inputs.python-version }}
-          environment-file: .github/requirements/conda-env-macOS-ARM64
-          pip-requirements-file: .github/requirements/pip-requirements-macOS.txt
-          default-packages: ""
+          environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}
+          pip-requirements-file: .github/requirements/pip-requirements-${{ runner.os }}.txt
 
       - name: Install sccache (only for non-forked PRs, and pushes to trunk)
         uses: nick-fields/retry@7152eba30c6575329ac0576536151aca5a72780e # v3.0.0

.github/workflows/_mac-test.yml

@@ -21,7 +21,7 @@ on:
       python-version:
         required: false
         type: string
-        default: "3.12"
+        default: "3.9"
         description: |
           The python version to be used. Will be 3.9 by default
       timeout-minutes:
@@ -144,9 +144,8 @@ jobs:
         uses: pytorch/test-infra/.github/actions/setup-miniconda@main
         with:
           python-version: ${{ inputs.python-version }}
-          environment-file: .github/requirements/conda-env-macOS-ARM64
-          pip-requirements-file: .github/requirements/pip-requirements-macOS.txt
-          default-packages: ""
+          environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}
+          pip-requirements-file: .github/requirements/pip-requirements-${{ runner.os }}.txt
 
       - name: Parse ref
         id: parse-ref
@@ -279,7 +278,6 @@ jobs:
           workflow_name: ${{ github.workflow }}
           workflow_run_id: ${{github.run_id}}
           workflow_attempt: ${{github.run_attempt}}
-          local_path: usage_log.txt
 
       - name: Clean up disk space
         if: always()

.github/workflows/_win-build.yml

@@ -23,7 +23,7 @@ on:
       vc-year:
         required: false
         type: string
-        default: "2022"
+        default: "2019"
         description: The Visual Studio year to use for building.
       build-with-debug:
         required: false
@@ -98,7 +98,7 @@ jobs:
             To start build locally, change working folder to \actions-runner\_work\pytorch\pytorch,
             Activate miniconda and Visual Studio environment, by running:
               call C:\Jenkins\Miniconda3\Scripts\activate.bat C:\Jenkins\Miniconda3
-              call "C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\Build\vcvarsall.bat" x64
+              call "C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC\Auxiliary\Build\vcvarsall.bat" x64
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch

.github/workflows/_win-test.yml

@@ -91,7 +91,7 @@ jobs:
             To start tests locally, change working folder to \actions-runner\_work\pytorch\pytorch\test,
             Activate miniconda and Visual Studio environment and set PYTHON_PATH, by running:
               call C:\Jenkins\Miniconda3\Scripts\activate.bat C:\Jenkins\Miniconda3
-              call "C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\Build\vcvarsall.bat" x64
+              call "C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC\Auxiliary\Build\vcvarsall.bat" x64
               set PYTHONPATH=C:\actions-runner\_work\pytorch\pytorch\build\win_tmp\build
 
       # [see note: pytorch repo ref]
@@ -191,8 +191,8 @@ jobs:
           NO_TD: ${{ steps.keep-going.outputs.ci-no-td }}
           VC_PRODUCT: "BuildTools"
           VC_VERSION: ""
-          VS_VERSION: "17.4.1"
-          VC_YEAR: "2022"
+          VS_VERSION: "16.8.6"
+          VC_YEAR: "2019"
           AWS_DEFAULT_REGION: us-east-1
           PR_NUMBER: ${{ github.event.pull_request.number }}
           GITHUB_REPOSITORY: ${{ github.repository }}

.github/workflows/assigntome-docathon.yml

@@ -28,14 +28,14 @@ jobs:
                   repo: context.repo.repo,
                   issue_number: issueNumber
                 });
-              const hasLabel = issue.labels.some(label => label.name === 'docathon-h1-2025');
+              const hasLabel = issue.labels.some(label => label.name === 'docathon-h1-2024');
               if (hasLabel) {
                 if (issue.assignee !== null) {
                   await github.rest.issues.createComment({
                     owner: context.repo.owner,
                     repo: context.repo.repo,
                     issue_number: issueNumber,
-                    body: "The issue is already assigned. Please pick an opened and unnasigned issue with the [docathon-h1-2025 label](https://github.com/pytorch/pytorch/issues?q=is%3Aopen+is%3Aissue+label%3Adocathon-h1-2025)."
+                    body: "The issue is already assigned. Please pick an opened and unnasigned issue with the [docathon-h1-2024 label](https://github.com/pytorch/pytorch/issues?q=is%3Aopen+is%3Aissue+label%3Adocathon-h1-2024)."
                   });
                 } else {
                   await github.rest.issues.addAssignees({
@@ -46,7 +46,7 @@ jobs:
                   });
                 }
               } else {
-                const commmentMessage = "This issue does not have the correct label. Please pick an opened and unnasigned issue with the [docathon-h1-2025 label](https://github.com/pytorch/pytorch/issues?q=is%3Aopen+is%3Aissue+label%3Adocathon-h1-2025)."
+                const commmentMessage = "This issue does not have the correct label. Please pick an opened and unnasigned issue with the [docathon-h1-2024 label](https://github.com/pytorch/pytorch/issues?q=is%3Aopen+is%3Aissue+label%3Adocathon-h1-2024)."
                 await github.rest.issues.createComment({
                   owner: context.repo.owner,
                   repo: context.repo.repo,

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

@@ -50,6 +50,7 @@ jobs:
         include: [
           { tag: "cuda12.8" },
           { tag: "cuda12.6" },
+          { tag: "cuda12.4" },
           { tag: "cuda11.8" },
           { tag: "rocm6.3"  },
           { tag: "rocm6.4"  },

.github/workflows/build-magma-windows.yml

@@ -19,15 +19,14 @@ concurrency:
 jobs:
   build-windows-magma:
     if: github.repository_owner == 'pytorch'
-    runs-on: windows-2022
+    runs-on: windows-2019
     strategy:
       matrix:
-        cuda_version: ["128", "126"]
+        cuda_version: ["128", "126", "118"]
         config: ["Release", "Debug"]
     env:
       CUDA_VERSION: ${{ matrix.cuda_version }}
       CONFIG: ${{ matrix.config }}
-      VC_YEAR: "2022"
     steps:
       - name: Checkout pytorch/pytorch
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

.github/workflows/build-manywheel-images-s390x.yml

@@ -3,9 +3,19 @@ name: Build manywheel docker images for s390x
 on:
   workflow_dispatch:
   push:
+    branches:
+      - main
+      - release/*
     tags:
-      - ciflow/s390/*
+      # NOTE: Binary build pipelines should only get triggered on release candidate or nightly builds
+      # Release candidate tags look like: v1.11.0-rc1
+      - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
     paths:
+      - .ci/docker/**
+      - .github/workflows/build-manywheel-images-s390x.yml
+  pull_request:
+    paths:
+      - .ci/docker/**
       - .github/workflows/build-manywheel-images-s390x.yml
 
 

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

@@ -49,6 +49,7 @@ jobs:
         include: [
           { name: "manylinux2_28-builder",          tag: "cuda12.8",          runner: "linux.9xlarge.ephemeral" },
           { name: "manylinux2_28-builder",          tag: "cuda12.6",          runner: "linux.9xlarge.ephemeral" },
+          { name: "manylinux2_28-builder",          tag: "cuda12.4",          runner: "linux.9xlarge.ephemeral" },
           { name: "manylinux2_28-builder",          tag: "cuda11.8",          runner: "linux.9xlarge.ephemeral" },
           { name: "manylinuxaarch64-builder",       tag: "cuda12.8",          runner: "linux.arm64.2xlarge.ephemeral" },
           { name: "manylinux2_28-builder",          tag: "rocm6.3",           runner: "linux.9xlarge.ephemeral" },

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

@@ -139,15 +139,6 @@ jobs:
 
           docker exec -t "${container_name}" yum install -y zlib-devel zip
           docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}"  -m pip install -U setuptools==78.1.0 pybind11==2.13.1 auditwheel wheel
-          set +e
-          docker exec -t "${container_name}" command -v pip
-          has_pip=$?
-          set -e
-          if [ $has_pip -eq 0 ] ; then
-              docker exec -t "${container_name}" pip install -U cmake --force-reinstall
-          else
-              docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}"  -m pip install -U cmake --force-reinstall
-          fi
 
           if [[ ("${{ matrix.device }}" == "cuda" || "${{ matrix.device }}" == "rocm" || "${{ matrix.device }}" == "aarch64" ) ]]; then
             # With this install, it gets clang 16.0.6.

.github/workflows/docker-builds.yml

@@ -49,21 +49,21 @@ jobs:
       matrix:
         runner: [linux.12xlarge]
         docker-image-name: [
-          pytorch-linux-jammy-cuda12.6-cudnn9-py3-gcc11,
-          pytorch-linux-jammy-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks,
-          pytorch-linux-jammy-cuda12.6-cudnn9-py3.12-gcc9-inductor-benchmarks,
-          pytorch-linux-jammy-cuda12.6-cudnn9-py3.13-gcc9-inductor-benchmarks,
-          pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks,
-          pytorch-linux-jammy-cuda12.8-cudnn9-py3.12-gcc9-inductor-benchmarks,
-          pytorch-linux-jammy-cuda12.8-cudnn9-py3.13-gcc9-inductor-benchmarks,
-          pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9,
-          pytorch-linux-jammy-py3.9-clang12,
-          pytorch-linux-jammy-py3.11-clang12,
-          pytorch-linux-jammy-py3.12-clang12,
-          pytorch-linux-jammy-py3.13-clang12,
+          pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9-inductor-benchmarks,
+          pytorch-linux-focal-cuda12.4-cudnn9-py3.12-gcc9-inductor-benchmarks,
+          pytorch-linux-focal-cuda12.4-cudnn9-py3.13-gcc9-inductor-benchmarks,
+          pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc11,
+          pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks,
+          pytorch-linux-focal-cuda12.6-cudnn9-py3.12-gcc9-inductor-benchmarks,
+          pytorch-linux-focal-cuda12.6-cudnn9-py3.13-gcc9-inductor-benchmarks,
+          pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9,
+          pytorch-linux-focal-py3.9-clang10,
+          pytorch-linux-focal-py3.11-clang10,
+          pytorch-linux-focal-py3.12-clang10,
+          pytorch-linux-focal-py3.13-clang10,
           pytorch-linux-jammy-rocm-n-1-py3,
           pytorch-linux-jammy-rocm-n-py3,
-          pytorch-linux-jammy-cuda12.8-cudnn9-py3.9-clang12,
+          pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-clang12,
           pytorch-linux-jammy-py3.9-gcc11,
           pytorch-linux-jammy-py3.9-gcc11-inductor-benchmarks,
           pytorch-linux-jammy-py3.12-halide,
@@ -71,9 +71,9 @@ jobs:
           pytorch-linux-jammy-xpu-2025.1-py3,
           pytorch-linux-jammy-py3-clang15-asan,
           pytorch-linux-jammy-py3-clang18-asan,
-          pytorch-linux-jammy-py3-clang12-onnx,
-          pytorch-linux-jammy-linter,
-          pytorch-linux-jammy-cuda12.8-cudnn9-py3.9-linter,
+          pytorch-linux-focal-py3-clang10-onnx,
+          pytorch-linux-focal-linter,
+          pytorch-linux-jammy-cuda11.8-cudnn9-py3.9-linter,
           pytorch-linux-jammy-py3-clang12-executorch,
           pytorch-linux-jammy-py3.12-triton-cpu
         ]
@@ -110,6 +110,18 @@ jobs:
           always-rebuild: true
           push: true
 
+      - name: Push docker image to old name
+        shell: bash
+        env:
+          ECR_DOCKER_IMAGE: ${{ steps.build-docker-image.outputs.docker-image }}
+        run: |
+          # This can be deleted after people have rebased their PRs/the new name has been in main for a while
+          set -euox pipefail
+          docker_image_name="308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/${{ matrix.docker-image-name }}"
+          foldersha=${ECR_DOCKER_IMAGE##*-}
+          docker tag "${ECR_DOCKER_IMAGE}" "${docker_image_name}:${foldersha}"
+          docker push "${docker_image_name}:${foldersha}"
+
       - name: Pull docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
         with:

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

@@ -136,7 +136,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_9-cuda-aarch64-12_8
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
       timeout-minutes: 420
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
@@ -252,7 +252,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_10-cuda-aarch64-12_8
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
       timeout-minutes: 420
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
@@ -368,7 +368,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_11-cuda-aarch64-12_8
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
       timeout-minutes: 420
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
@@ -484,7 +484,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_12-cuda-aarch64-12_8
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
       timeout-minutes: 420
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
@@ -600,7 +600,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_13-cuda-aarch64-12_8
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
       timeout-minutes: 420
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
@@ -716,7 +716,7 @@ jobs:
       ALPINE_IMAGE: "arm64v8/alpine"
       build_name: manywheel-py3_13t-cuda-aarch64-12_8
       build_environment: linux-aarch64-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
       timeout-minutes: 420
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}

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

@@ -108,7 +108,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_6-test:  # Testing
@@ -155,7 +155,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_8-test:  # Testing

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

@@ -200,7 +200,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_6-test:  # Testing
@@ -269,7 +269,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_9-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_9-cuda12_8-test:  # Testing
@@ -813,7 +813,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_10-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_10-cuda12_6-test:  # Testing
@@ -882,7 +882,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_10-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_10-cuda12_8-test:  # Testing
@@ -1426,7 +1426,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_11-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_11-cuda12_6-test:  # Testing
@@ -1563,7 +1563,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_11-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_11-cuda12_8-test:  # Testing
@@ -2107,7 +2107,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_12-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_12-cuda12_6-test:  # Testing
@@ -2176,7 +2176,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_12-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_12-cuda12_8-test:  # Testing
@@ -2720,7 +2720,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_13-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_13-cuda12_6-test:  # Testing
@@ -2789,7 +2789,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_13-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_13-cuda12_8-test:  # Testing
@@ -3333,7 +3333,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_13t-cuda12_6
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.6.80; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.5.1.17; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.6.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.0.4; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.7.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.1.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.4.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.6.77; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.6.85; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.11.1.6; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_13t-cuda12_6-test:  # Testing
@@ -3402,7 +3402,7 @@ jobs:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
       build_name: manywheel-py3_13t-cuda12_8
       build_environment: linux-binary-manywheel
-      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.4.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.83; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.3.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.7.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvshmem-cu12==3.2.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.90; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.93; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.1.3; platform_system == 'Linux' and platform_machine == 'x86_64'
+      PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.8.57; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.8.0.87; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.8.3.14; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.3.3.41; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.9.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.7.2.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.5.7.53; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.26.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.8.55; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.8.61; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufile-cu12==1.13.0.11; platform_system == 'Linux' and platform_machine == 'x86_64'
     secrets:
       github-token: ${{ secrets.GITHUB_TOKEN }}
   manywheel-py3_13t-cuda12_8-test:  # Testing

.github/workflows/generated-windows-arm64-binary-libtorch-debug-nightly.yml

@@ -50,7 +50,7 @@ jobs:
   libtorch-cpu-shared-with-deps-debug-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -77,12 +77,23 @@ jobs:
         run: |
           mkdir "%NIGHTLIES_PYTORCH_ROOT%"
           mkdir "%PYTORCH_FINAL_PACKAGE_DIR%"
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Bootstrap Build Tools
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
-      - name: Git checkout PyTorch
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
+      - name: Bootstrap Git
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
+      - name: Git checkout PyTorch - recursive
         uses: actions/checkout@v4
         with:
           path: "pytorch"
@@ -127,7 +138,7 @@ jobs:
     needs:
       - libtorch-cpu-shared-with-deps-debug-build
       - get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -149,11 +160,18 @@ jobs:
           echo BINARY_ENV_FILE=%RUNNER_TEMP%/env>> %GITHUB_ENV%
           echo PYTORCH_FINAL_PACKAGE_DIR=%RUNNER_TEMP%/artifacts>> %GITHUB_ENV%
           echo WIN_PACKAGE_WORK_DIR=%RUNNER_TEMP%>> %GITHUB_ENV%
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Populate binary env
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
       - name: Git checkout PyTorch
         uses: actions/checkout@v4
         with:
@@ -167,6 +185,10 @@ jobs:
         shell: cmd
         run: |
           "pytorch/.ci/pytorch/windows/arm64/bootstrap_python.bat"
+      - name: Bootstrap Build Tools
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
       - name: Bootstrap Rust
         shell: cmd
         run: |

.github/workflows/generated-windows-arm64-binary-libtorch-release-nightly.yml

@@ -50,7 +50,7 @@ jobs:
   libtorch-cpu-shared-with-deps-release-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -77,12 +77,23 @@ jobs:
         run: |
           mkdir "%NIGHTLIES_PYTORCH_ROOT%"
           mkdir "%PYTORCH_FINAL_PACKAGE_DIR%"
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Bootstrap Build Tools
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
-      - name: Git checkout PyTorch
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
+      - name: Bootstrap Git
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
+      - name: Git checkout PyTorch - recursive
         uses: actions/checkout@v4
         with:
           path: "pytorch"
@@ -127,7 +138,7 @@ jobs:
     needs:
       - libtorch-cpu-shared-with-deps-release-build
       - get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -149,11 +160,18 @@ jobs:
           echo BINARY_ENV_FILE=%RUNNER_TEMP%/env>> %GITHUB_ENV%
           echo PYTORCH_FINAL_PACKAGE_DIR=%RUNNER_TEMP%/artifacts>> %GITHUB_ENV%
           echo WIN_PACKAGE_WORK_DIR=%RUNNER_TEMP%>> %GITHUB_ENV%
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Populate binary env
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
       - name: Git checkout PyTorch
         uses: actions/checkout@v4
         with:
@@ -167,6 +185,10 @@ jobs:
         shell: cmd
         run: |
           "pytorch/.ci/pytorch/windows/arm64/bootstrap_python.bat"
+      - name: Bootstrap Build Tools
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
       - name: Bootstrap Rust
         shell: cmd
         run: |

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

@@ -50,7 +50,7 @@ jobs:
   wheel-py3_11-cpu-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -73,12 +73,23 @@ jobs:
         run: |
           mkdir "%NIGHTLIES_PYTORCH_ROOT%"
           mkdir "%PYTORCH_FINAL_PACKAGE_DIR%"
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Bootstrap Build Tools
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
-      - name: Git checkout PyTorch
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
+      - name: Bootstrap Git
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
+      - name: Git checkout PyTorch - recursive
         uses: actions/checkout@v4
         with:
           path: "pytorch"
@@ -123,7 +134,7 @@ jobs:
     needs:
       - wheel-py3_11-cpu-build
       - get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -141,11 +152,18 @@ jobs:
           echo BINARY_ENV_FILE=%RUNNER_TEMP%/env>> %GITHUB_ENV%
           echo PYTORCH_FINAL_PACKAGE_DIR=%RUNNER_TEMP%/artifacts>> %GITHUB_ENV%
           echo WIN_PACKAGE_WORK_DIR=%RUNNER_TEMP%>> %GITHUB_ENV%
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Populate binary env
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
       - name: Git checkout PyTorch
         uses: actions/checkout@v4
         with:
@@ -159,6 +177,10 @@ jobs:
         shell: cmd
         run: |
           "pytorch/.ci/pytorch/windows/arm64/bootstrap_python.bat"
+      - name: Bootstrap Build Tools
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
       - name: Bootstrap Rust
         shell: cmd
         run: |
@@ -197,7 +219,7 @@ jobs:
   wheel-py3_12-cpu-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -220,12 +242,23 @@ jobs:
         run: |
           mkdir "%NIGHTLIES_PYTORCH_ROOT%"
           mkdir "%PYTORCH_FINAL_PACKAGE_DIR%"
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Bootstrap Build Tools
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
-      - name: Git checkout PyTorch
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
+      - name: Bootstrap Git
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
+      - name: Git checkout PyTorch - recursive
         uses: actions/checkout@v4
         with:
           path: "pytorch"
@@ -270,7 +303,7 @@ jobs:
     needs:
       - wheel-py3_12-cpu-build
       - get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -288,11 +321,18 @@ jobs:
           echo BINARY_ENV_FILE=%RUNNER_TEMP%/env>> %GITHUB_ENV%
           echo PYTORCH_FINAL_PACKAGE_DIR=%RUNNER_TEMP%/artifacts>> %GITHUB_ENV%
           echo WIN_PACKAGE_WORK_DIR=%RUNNER_TEMP%>> %GITHUB_ENV%
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Populate binary env
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
       - name: Git checkout PyTorch
         uses: actions/checkout@v4
         with:
@@ -306,6 +346,10 @@ jobs:
         shell: cmd
         run: |
           "pytorch/.ci/pytorch/windows/arm64/bootstrap_python.bat"
+      - name: Bootstrap Build Tools
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
       - name: Bootstrap Rust
         shell: cmd
         run: |
@@ -344,7 +388,7 @@ jobs:
   wheel-py3_13-cpu-build:
     if: ${{ github.repository_owner == 'pytorch' }}
     needs: get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -367,12 +411,23 @@ jobs:
         run: |
           mkdir "%NIGHTLIES_PYTORCH_ROOT%"
           mkdir "%PYTORCH_FINAL_PACKAGE_DIR%"
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Bootstrap Build Tools
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
-      - name: Git checkout PyTorch
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
+      - name: Bootstrap Git
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
+      - name: Git checkout PyTorch - recursive
         uses: actions/checkout@v4
         with:
           path: "pytorch"
@@ -417,7 +472,7 @@ jobs:
     needs:
       - wheel-py3_13-cpu-build
       - get-label-type
-    runs-on: "windows-11-arm64-preview"
+    runs-on: "windows-11-arm64"
     timeout-minutes: 300
     env:
       PYTORCH_ROOT: ${{ github.workspace }}/pytorch
@@ -435,11 +490,18 @@ jobs:
           echo BINARY_ENV_FILE=%RUNNER_TEMP%/env>> %GITHUB_ENV%
           echo PYTORCH_FINAL_PACKAGE_DIR=%RUNNER_TEMP%/artifacts>> %GITHUB_ENV%
           echo WIN_PACKAGE_WORK_DIR=%RUNNER_TEMP%>> %GITHUB_ENV%
-      - name: Enable long paths
+      - name: Git checkout PyTorch
+        uses: actions/checkout@v4
+        with:
+          path: "pytorch"
+      - name: Populate binary env
         shell: cmd
         run: |
-          git config --system --get core.longpaths || echo "core.longpaths is not set, setting it now"
-          git config --system core.longpaths true
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_git.bat"
+      - name: Remove Pytorch folder
+        shell: cmd
+        run: |
+          rmdir /s /q "pytorch"
       - name: Git checkout PyTorch
         uses: actions/checkout@v4
         with:
@@ -453,6 +515,10 @@ jobs:
         shell: cmd
         run: |
           "pytorch/.ci/pytorch/windows/arm64/bootstrap_python.bat"
+      - name: Bootstrap Build Tools
+        shell: cmd
+        run: |
+          "pytorch/.ci/pytorch/windows/arm64/bootstrap_buildtools.bat"
       - name: Bootstrap Rust
         shell: cmd
         run: |

.github/workflows/h100-distributed.yml

@@ -1,53 +0,0 @@
-name: Limited CI for distributed tests on H100
-
-on:
-  pull_request:
-    paths:
-      - .github/workflows/h100-distributed.yml
-  workflow_dispatch:
-  push:
-    tags:
-      - ciflow/h100-distributed/*
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}-${{ github.event_name == 'schedule' }}
-  cancel-in-progress: true
-
-jobs:
-
-  get-label-type:
-    if: github.repository_owner == 'pytorch'
-    name: get-label-type
-    uses: pytorch/pytorch/.github/workflows/_runner-determinator.yml@main
-    with:
-      triggering_actor: ${{ github.triggering_actor }}
-      issue_owner: ${{ github.event.pull_request.user.login || github.event.issue.user.login }}
-      curr_branch: ${{ github.head_ref || github.ref_name }}
-      curr_ref_type: ${{ github.ref_type }}
-
-  linux-focal-cuda12_6-py3_10-gcc11-sm90-build:
-    name: linux-focal-cuda12.6-py3.10-gcc11-sm90
-    uses: ./.github/workflows/_linux-build.yml
-    needs: get-label-type
-    with:
-      runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      runner: "linux.12xlarge"
-      build-environment: linux-focal-cuda12.6-py3.10-gcc11-sm90
-      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc11
-      cuda-arch-list: '9.0'
-      test-matrix: |
-        { include: [
-          { config: "h100_distributed", shard: 1, num_shards: 1, runner: "linux.aws.h100.8" },
-        ]}
-    secrets: inherit
-
-  linux-focal-cuda12_6-py3_10-gcc11-sm90-test:
-    name: linux-focal-cuda12.6-py3.10-gcc11-sm90
-    uses: ./.github/workflows/_linux-test.yml
-    needs:
-      - linux-focal-cuda12_6-py3_10-gcc11-sm90-build
-    with:
-      build-environment: linux-focal-cuda12.6-py3.10-gcc11-sm90
-      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc11-sm90-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc11-sm90-build.outputs.test-matrix }}
-    secrets: inherit

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

@@ -27,15 +27,15 @@ jobs:
       curr_ref_type: ${{ github.ref_type }}
       opt_out_experiments: lf
 
-  build:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-micro-benchmark-build:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-build.yml
     needs:
       - get-default-label-prefix
     with:
       runner_prefix: "${{ needs.get-default-label-prefix.outputs.label-type }}"
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.0'
       test-matrix: |
         { include: [
@@ -43,13 +43,13 @@ jobs:
         ]}
     secrets: inherit
 
-  test:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-micro-benchmark-test:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-test.yml
-    needs: build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-micro-benchmark-build
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image: ${{ needs.build.outputs.docker-image }}
-      test-matrix: ${{ needs.build.outputs.test-matrix }}
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-micro-benchmark-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-micro-benchmark-build.outputs.test-matrix }}
       timeout-minutes: 720
     secrets: inherit

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

@@ -24,15 +24,15 @@ jobs:
       curr_ref_type: ${{ github.ref_type }}
       opt_out_experiments: lf
 
-  build:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-build:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-build.yml
     needs:
       - get-default-label-prefix
     with:
       runner_prefix: "${{ needs.get-default-label-prefix.outputs.label-type }}"
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.0'
       test-matrix: |
         { include: [
@@ -43,14 +43,14 @@ jobs:
         ]}
     secrets: inherit
 
-  test:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-test:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-test.yml
-    needs: build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-build
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image: ${{ needs.build.outputs.docker-image }}
-      test-matrix: ${{ needs.build.outputs.test-matrix }}
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.test-matrix }}
       # disable monitor in perf tests for more investigation
       disable-monitor: false
       monitor-log-interval: 15

.github/workflows/inductor-perf-test-nightly-h100.yml

@@ -79,13 +79,13 @@ jobs:
 
   # NB: Keep this in sync with trunk.yml
   build:
-    name: cuda12.8-py3.10-gcc9-sm90
+    name: cuda12.6-py3.10-gcc9-sm90
     uses: ./.github/workflows/_linux-build.yml
     needs: get-label-type
     with:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm90
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm90
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '9.0'
       test-matrix: |
         { include: [
@@ -111,12 +111,12 @@ jobs:
     secrets: inherit
 
   test-nightly:
-    name: cuda12.8-py3.10-gcc9-sm90
+    name: cuda12.6-py3.10-gcc9-sm90
     uses: ./.github/workflows/_linux-test.yml
     needs: build
     if: github.event.schedule == '0 7 * * 1-6'
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm90
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm90
       dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true-aotinductor-true-freezing_cudagraphs-true-cudagraphs_low_precision-true
       docker-image: ${{ needs.build.outputs.docker-image }}
       test-matrix: ${{ needs.build.outputs.test-matrix }}
@@ -128,12 +128,12 @@ jobs:
     secrets: inherit
 
   test-weekly:
-    name: cuda12.8-py3.10-gcc9-sm90
+    name: cuda12.6-py3.10-gcc9-sm90
     uses: ./.github/workflows/_linux-test.yml
     needs: build
     if: github.event.schedule == '0 7 * * 0'
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm90
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm90
       dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true-aotinductor-true-freezing_cudagraphs-true-maxautotune-true-freeze_autotune_cudagraphs-true-cudagraphs_low_precision-true
       docker-image: ${{ needs.build.outputs.docker-image }}
       test-matrix: ${{ needs.build.outputs.test-matrix }}
@@ -145,12 +145,12 @@ jobs:
     secrets: inherit
 
   test:
-    name: cuda12.8-py3.10-gcc9-sm90
+    name: cuda12.6-py3.10-gcc9-sm90
     uses: ./.github/workflows/_linux-test.yml
     needs: build
     if: github.event_name == 'workflow_dispatch'
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm90
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm90
       dashboard-tag: training-${{ inputs.training }}-inference-${{ inputs.inference }}-default-${{ inputs.default }}-dynamic-${{ inputs.dynamic }}-cudagraphs-${{ inputs.cudagraphs }}-cppwrapper-${{ inputs.cppwrapper }}-aotinductor-${{ inputs.aotinductor }}-maxautotune-${{ inputs.maxautotune }}-freezing_cudagraphs-${{ inputs.freezing_cudagraphs }}-cudagraphs_low_precision-${{ inputs.cudagraphs }}
       docker-image: ${{ needs.build.outputs.docker-image }}
       test-matrix: ${{ needs.build.outputs.test-matrix }}

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

@@ -42,7 +42,7 @@ jobs:
       runner-type: macos-m1-stable
       build-generates-artifacts: true
       # To match the one pre-installed in the m1 runners
-      python-version: 3.12.7
+      python-version: 3.9.12
       test-matrix: |
         { include: [
           { config: "perf_smoketest", shard: 1, num_shards: 1, runner: "macos-m2-15" },
@@ -56,10 +56,8 @@ jobs:
     with:
       build-environment: macos-py3-arm64-distributed
       # Same as the build job
-      python-version: 3.12.7
+      python-version: 3.9.12
       test-matrix: ${{ needs.macos-perf-py3-arm64-build.outputs.test-matrix }}
-      disable-monitor: false
-      monitor-log-interval: 15
-      monitor-data-collect-interval: 4
-
+      # disable monitor in perf tests for more investigation
+      disable-monitor: true
     secrets: inherit

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

@@ -78,14 +78,14 @@ jobs:
       opt_out_experiments: lf
 
   # NB: Keep this in sync with trunk.yml
-  build:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-build:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-build.yml
     needs: get-label-type
     with:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.0'
       test-matrix: |
         { include: [
@@ -112,32 +112,32 @@ jobs:
       selected-test-configs: ${{ inputs.benchmark_configs }}
     secrets: inherit
 
-  test-nightly:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-test-nightly:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-test.yml
-    needs: build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-build
     if: github.event.schedule == '0 7 * * 1-6'
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
       dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true-aotinductor-true-freezing_cudagraphs-true-cudagraphs_low_precision-true
-      docker-image: ${{ needs.build.outputs.docker-image }}
-      test-matrix: ${{ needs.build.outputs.test-matrix }}
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.test-matrix }}
       timeout-minutes: 720
       disable-monitor: false
       monitor-log-interval: 15
       monitor-data-collect-interval: 4
     secrets: inherit
 
-  test-weekly:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-test-weekly:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-test.yml
-    needs: build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-build
     if: github.event.schedule == '0 7 * * 0'
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
       dashboard-tag: training-true-inference-true-default-true-dynamic-true-cudagraphs-true-cppwrapper-true-aotinductor-true-freezing_cudagraphs-true-maxautotune-true-freeze_autotune_cudagraphs-true-cudagraphs_low_precision-true
-      docker-image: ${{ needs.build.outputs.docker-image }}
-      test-matrix: ${{ needs.build.outputs.test-matrix }}
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.test-matrix }}
       timeout-minutes: 1440
       # disable monitor in perf tests, next step is to enable it
       disable-monitor: false
@@ -145,16 +145,16 @@ jobs:
       monitor-data-collect-interval: 4
     secrets: inherit
 
-  test:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-test:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-test.yml
-    needs: build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-build
     if: github.event_name == 'workflow_dispatch'
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
       dashboard-tag: training-${{ inputs.training }}-inference-${{ inputs.inference }}-default-${{ inputs.default }}-dynamic-${{ inputs.dynamic }}-cudagraphs-${{ inputs.cudagraphs }}-cppwrapper-${{ inputs.cppwrapper }}-aotinductor-${{ inputs.aotinductor }}-maxautotune-${{ inputs.maxautotune }}-freezing_cudagraphs-${{ inputs.freezing_cudagraphs }}-cudagraphs_low_precision-${{ inputs.cudagraphs }}
-      docker-image: ${{ needs.build.outputs.docker-image }}
-      test-matrix: ${{ needs.build.outputs.test-matrix }}
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.test-matrix }}
       timeout-minutes: 720
       disable-monitor: false
       monitor-log-interval: 15

.github/workflows/inductor-periodic.yml

@@ -29,14 +29,14 @@ jobs:
       curr_ref_type: ${{ github.ref_type }}
       opt_out_experiments: lf
 
-  linux-jammy-cuda12_8-py3_10-gcc9-periodic-dynamo-benchmarks-build:
-    name: cuda12.8-py3.10-gcc9-sm86-periodic-dynamo-benchmarks
+  linux-focal-cuda12_6-py3_10-gcc9-periodic-dynamo-benchmarks-build:
+    name: cuda12.6-py3.10-gcc9-sm86-periodic-dynamo-benchmarks
     uses: ./.github/workflows/_linux-build.yml
     needs: get-default-label-prefix
     with:
       runner_prefix: "${{ needs.get-default-label-prefix.outputs.label-type }}"
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm86
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm86
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.6'
       test-matrix: |
         { include: [
@@ -58,14 +58,14 @@ jobs:
         ]}
     secrets: inherit
 
-  linux-jammy-cuda12_8-py3_10-gcc9-periodic-dynamo-benchmarks-test:
-    name: cuda12.8-py3.10-gcc9-sm86-periodic-dynamo-benchmarks
+  linux-focal-cuda12_6-py3_10-gcc9-periodic-dynamo-benchmarks-test:
+    name: cuda12.6-py3.10-gcc9-sm86-periodic-dynamo-benchmarks
     uses: ./.github/workflows/_linux-test.yml
-    needs: linux-jammy-cuda12_8-py3_10-gcc9-periodic-dynamo-benchmarks-build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-periodic-dynamo-benchmarks-build
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm86
-      docker-image: ${{ needs.linux-jammy-cuda12_8-py3_10-gcc9-periodic-dynamo-benchmarks-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-jammy-cuda12_8-py3_10-gcc9-periodic-dynamo-benchmarks-build.outputs.test-matrix }}
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm86
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-periodic-dynamo-benchmarks-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-periodic-dynamo-benchmarks-build.outputs.test-matrix }}
     secrets: inherit
 
   linux-jammy-rocm-py3_10-periodic-dynamo-benchmarks-build:
@@ -109,15 +109,15 @@ jobs:
       test-matrix: ${{ needs.linux-jammy-rocm-py3_10-periodic-dynamo-benchmarks-build.outputs.test-matrix }}
     secrets: inherit
 
-  linux-jammy-cuda12_8-py3_10-gcc9-inductor-smoke-build:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-build-gcp:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-build.yml
     needs:
       - get-default-label-prefix
     with:
       runner_prefix: "${{ needs.get-default-label-prefix.outputs.label-type }}"
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.0'
       test-matrix: |
         { include: [
@@ -125,14 +125,14 @@ jobs:
         ]}
     secrets: inherit
 
-  linux-jammy-cuda12_8-py3_10-gcc9-inductor-smoke-test:
-    name: cuda12.8-py3.10-gcc9-sm80
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-test-gcp:
+    name: cuda12.6-py3.10-gcc9-sm80
     uses: ./.github/workflows/_linux-test.yml
-    needs: linux-jammy-cuda12_8-py3_10-gcc9-inductor-smoke-build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-build-gcp
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm80
-      docker-image: ${{ needs.linux-jammy-cuda12_8-py3_10-gcc9-inductor-smoke-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-jammy-cuda12_8-py3_10-gcc9-inductor-smoke-build.outputs.test-matrix }}
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm80
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build-gcp.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build-gcp.outputs.test-matrix }}
     secrets: inherit
 
   linux-jammy-cpu-py3_9-gcc11-periodic-dynamo-benchmarks-build:
@@ -170,16 +170,16 @@ jobs:
     secrets: inherit
 
 
-  linux-jammy-cuda12_8-py3_10-gcc9-inductor-build:
-    name: cuda12.8-py3.10-gcc9-sm86
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-build:
+    name: cuda12.6-py3.10-gcc9-sm86
     uses: ./.github/workflows/_linux-build.yml
     needs: get-default-label-prefix
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm86
-      docker-image-name: ci-image:pytorch-linux-jammy-cuda12.8-cudnn9-py3-gcc9-inductor-benchmarks
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm86
+      docker-image-name: ci-image:pytorch-linux-focal-cuda12.6-cudnn9-py3-gcc9-inductor-benchmarks
       cuda-arch-list: '8.6'
       runner_prefix: "${{ needs.get-default-label-prefix.outputs.label-type }}"
-      sync-tag: linux-jammy-cuda12_8-py3_10-gcc9-inductor-build
+      sync-tag: linux-focal-cuda12_6-py3_10-gcc9-inductor-build
       test-matrix: |
         { include: [
           { config: "dynamic_inductor_huggingface", shard: 1, num_shards: 1, runner: "linux.g5.4xlarge.nvidia.gpu" },
@@ -195,14 +195,14 @@ jobs:
         ]}
     secrets: inherit
 
-  linux-jammy-cuda12_8-py3_10-gcc9-inductor-test:
-    name: cuda12.8-py3.10-gcc9-sm86
+  linux-focal-cuda12_6-py3_10-gcc9-inductor-test:
+    name: cuda12.6-py3.10-gcc9-sm86
     uses: ./.github/workflows/_linux-test.yml
-    needs: linux-jammy-cuda12_8-py3_10-gcc9-inductor-build
+    needs: linux-focal-cuda12_6-py3_10-gcc9-inductor-build
     with:
-      build-environment: linux-jammy-cuda12.8-py3.10-gcc9-sm86
-      docker-image: ${{ needs.linux-jammy-cuda12_8-py3_10-gcc9-inductor-build.outputs.docker-image }}
-      test-matrix: ${{ needs.linux-jammy-cuda12_8-py3_10-gcc9-inductor-build.outputs.test-matrix }}
+      build-environment: linux-focal-cuda12.6-py3.10-gcc9-sm86
+      docker-image: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.docker-image }}
+      test-matrix: ${{ needs.linux-focal-cuda12_6-py3_10-gcc9-inductor-build.outputs.test-matrix }}
     secrets: inherit
 
   linux-jammy-cpu-py3_9-gcc11-inductor-build:

.github/workflows/inductor-rocm-mi300.yml

@@ -38,12 +38,12 @@ jobs:
       opt_out_experiments: lf
 
   linux-jammy-rocm-py3_10-inductor-build:
-    name: rocm-py3.10-inductor-mi300
+    name: rocm-py3.10-inductor
     uses: ./.github/workflows/_linux-build.yml
     needs: get-label-type
     with:
       runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
-      build-environment: linux-jammy-rocm-py3.10-mi300
+      build-environment: linux-jammy-rocm-py3.10
       docker-image-name: ci-image:pytorch-linux-jammy-rocm-n-py3
       test-matrix: |
         { include: [
@@ -56,11 +56,11 @@ jobs:
     permissions:
       id-token: write
       contents: read
-    name: rocm-py3.10-inductor-mi300
+    name: rocm-py3.10-inductor
     uses: ./.github/workflows/_rocm-test.yml
     needs: linux-jammy-rocm-py3_10-inductor-build
     with:
-      build-environment: linux-jammy-rocm-py3.10-mi300
+      build-environment: linux-jammy-rocm-py3.10
       docker-image: ${{ needs.linux-jammy-rocm-py3_10-inductor-build.outputs.docker-image }}
       test-matrix:  ${{ needs.linux-jammy-rocm-py3_10-inductor-build.outputs.test-matrix }}
     secrets: inherit

.github/workflows/inductor-rocm.yml

@@ -43,7 +43,6 @@ jobs:
           { config: "inductor", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.2" },
           { config: "inductor", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.2" },
         ]}
-      allow-reuse-old-whl: true
     secrets: inherit
 
   linux-jammy-rocm-py3_10-inductor-test: